longtable leaves final page with default margins

A multi-page table formats more or less OK, but the last page reverts to the default margins.

(Minor question: I assume my inability to make the table fill the height of the page with \textheight is a noob error.)

Code example shortened from 15 pages, still too long:

\documentclass[letterpaper,10pt,notitlepage]{report}
%%\usepackage{fullpage}
\usepackage[utf8x]{inputenc}
\usepackage{longtable}

\begin{document}
\title{\textit{mnp} Model}
\author{ Gregg Hauser}
\maketitle

\begin{abstract}
\end{abstract}
\begingroup
% lousy job of formatting the page - heading and page number are way left of center ...
\setlength{\hoffset}{-1.25in}
\setlength{\voffset}{-.5in}
\setlength{\topmargin}{0pt}
\setlength{\textheight}{9in}
%
%chapter Unsolved Problems in Physics
\textit{mnp} Model and the Unsolved Problems in Physics\\
http:\slash \slash en.wikipedia.org\slash wiki\slash List\_of\_unsolved\_problems\_in\_physics\\
\\
\begingroup
\newskip\LTpre \LTpre=0pt
\newskip\LTpost \LTpost=0pt
\begin{longtable}{ p{1in} p{.7in} p{3.5in} p{1.5in} }
%%\hline
% use & to separate \\ \hline to end
Category\slash Issue & Date & Comments & Confidence \\ \hline
\multicolumn{4}{c}%
{\tablename\ \thetable\ -- \textit{Continued from previous page}} \\ %% \hline
Category\slash Issue & Date & Comments & Confidence \\ \hline
\\
\multicolumn{4}{l}{\textbf{Quantum gravity, cosmology, general relativity}} \\ %% \hline
\raggedright Vacuum catastrophe & 2011-11-05 & 10\^14 GV\slash m3 as vacuum energy density (higher than other numbers?) -
does that suggest that a figment has a known size? In the \textit{mnp} Model a cubic meter of space could be empty
or could have an integer number of figments. Gravity (m-figments) would still work through such a region
(no gravitational effects in truly empty regions, though). Particles might “stop” but would be bringing in figments,
so will not stay in place. In “our neighborhood”, if each
(calculated) 1x10\^-9erg per cm\^3 or 10\^-16J or 6.2415x10\^2eV per cm\^3 or 6.2415x10\^11eV\slash m\^3 & low \tabularnewline %% \hline
Quantum Gravity & 2011-11-05 & The nature of the \textit{mnp} Model.
The model suggests that objects appear lighter to heavier than terrestrial gravity would suggest as follows:
black holes appear lighter (and lighter yet from other galaxies),
terrestrial objects appear “normal” at intermediate distances but lighter at inter galactic distances,
emissive objects like the sun appear heavier than terrestrial objects at all distances,
but may appear “lighter” as the distance increases,
and massively emissive objects such as super novas appear much more massive than they (were).
Gravity is propagated at the speed of light (and less). & high \\ %% \hline
Black Holes & 2011-11-05 & The “thermal” radiation emitted by black holes is m-figments,
which are emitted (and need to be emitted so that black holes participate in gravitation).
“Information” in the number of m, n, and p figments remains. Structures remain,
though the \textit{mnp} Model sees more rearrangements of structure than particle physics does.
Neutrinos may pass through black holes, though greater contents may reduce the odds of transit.
The neutrinos emerging would be mediators of (long distance) gravity as well. & high \\ %% \hline
\raggedright Quasiparticles and Phonons & 2011-10-29 & Looking at the Wikipedia phonon article,
the suggestion that Fourier analysis proves quantum effects strikes the author as a tautology.
A finite Fourier transform on a finite field will only yield wavelengths as integral multiples
(or rational fractions at best) which when inverted will leave continuities out of the picture.
Fourier analysis is very useful for calculation and maybe for understanding "points of view" and possibly for finding valleys or peaks in potential.
The author is not comfortable with discrete analysis creating quasi particles that then interact with electrons.
Maybe better left unsaid in professional company. & medium low
\tabularnewline %% \hline

\end{longtable}
\endgroup

unreviewed 2012-08-12
\endgroup

\end{document}

-
Welcome to TeX.SE. You should add \usepacakge{showframe} to see that you are way past the margins on the page. –  Peter Grill Aug 13 '12 at 1:36
showpage is useful - I can get the boundaries "right enough" within the group, but the last page still reverts. If I bite the bullet and accept that the page description for long tables has to match the global description, things just work. Minor questions: So groups do not nest in tex the way environments do? And/or is longtable postponing some typesetting long beyond \end{longtable} ? –  Gregg Hauser Aug 13 '12 at 22:26
@DavidCarlisle \hoffset and \voffset can also be used after the first page. –  Heiko Oberdiek Aug 14 '12 at 14:36
@GreggHauser The end of group is in the middle of the page. LaTeX constructs the page in the output routine that is called later if there is enough material to complete a page. But then the page layout parameters are already reverted. End the page (\newpage or similar) before you end the group. –  Heiko Oberdiek Aug 14 '12 at 14:42
@DavidCarlisle Sorry, they are not even global. The following plain TeX file locally changes \hoffset in a group: Page 1\par \begingroup \hoffset=10cm \vfill \eject \endgroup Page 2 \bye –  Heiko Oberdiek Aug 14 '12 at 14:53

Margins
You have put the margin changes within a begingroup, so it is terminated when the group end.

If your purpose with the begin{group} is to change margin for a part of the document, there is several better options at CTAN. See this How can I change the margins for only part of the text? question for several, different possibilities.

If you need really small margins throughout the documents, I suggest hat you load the package typearea with a really big DIV-factor:

\usepackage[DIV=20]{typearea}


makes the typeblock wide enough for your table. _You may narrow the margin for the start of the document by putting that part within a addmargin-environment.

The pages is not filled up because longtable does not break a row between pages. You have som cell with lots of text, and those will be push to the next page. Either you have to reduce the text, or manually split the table yourself.

LTleft
To only move the table to the left, without changing margin for the whole document, you can try using these commands, which I found by reference to Ulrike Fisher’s answer on comp.text.tex, see How to change the margins for a landscape-orientated longtable in classicthesis, updated with Heiko Oberdiek’s suggestion (ref. comment below) :

\setlength{\LTleft}{0pt plus 1fill minus 1fill}%
\setlength{\LTright}{\LTleft}


This code centres the table between the margin.

New MWE
I have made a new MWE from your code. I have added some additional packages, url for nicer links, booktabs for better rules in the tables, including better spacing between rows. Also, I suggest that you use @{} in the start an end of the column definition to get rid of the overhang:

\documentclass[letterpaper,10pt,notitlepage]{report}
\usepackage[utf8x]{inputenc}
\usepackage{longtable}
\usepackage[DIV=9]{typearea}
\usepackage{scrextend,booktabs}
\usepackage{blindtext,url}

\begin{document}
\title{\textit{mnp} Model}
\author{Gregg Hauser}
\maketitle
\begin{abstract}
\blindtext
\end{abstract}

%chapter Unsolved Problems in Physics
\noindent\textit{mnp} Model and the Unsolved Problems in Physics\\
\url{http:\\en.wikipedia.org\wiki\List_of_unsolved_problems_in_physics}

\setlength{\LTleft}{0pt plus 1fill minus 1fill}%
\setlength{\LTright}{\LTleft}

\begin{longtable}{@{} p{1in} p{.7in} p{3.5in} p{1.5in}@{}}
\toprule
% use & to separate \\ \hline to end
Category\slash Issue & Date & Comments & Confidence \\ \midrule
\multicolumn{4}{@{}c}%
{\tablename\ \thetable\ -- \textit{Continued from previous page}} \\ \midrule
Category\slash Issue & Date & Comments & Confidence \\ \midrule
\multicolumn{4}{@{}l}{\textbf{Quantum gravity, cosmology, general relativity}} \\\midrule
\raggedright Vacuum catastrophe & 2011-11-05 & 10\^14 GV\slash m3 as vacuum energy density (higher than other numbers?) -
does that suggest that a figment has a known size? In the \textit{mnp} Model a cubic meter of space could be empty
or could have an integer number of figments. Gravity (m-figments) would still work through such a region
(no gravitational effects in truly empty regions, though). Particles might “stop” but would be bringing in figments,
so will not stay in place. In “our neighborhood”, if each
(calculated) 1x10\^-9erg per cm\^3 or 10\^-16J or 6.2415x10\^2eV per cm\^3 or 6.2415x10\^11eV\slash m\^3 & low \tabularnewline  \midrule
Quantum Gravity & 2011-11-05 & The nature of the \textit{mnp} Model.
The model suggests that objects appear lighter to heavier than terrestrial gravity would suggest as follows:
black holes appear lighter (and lighter yet from other galaxies),
terrestrial objects appear “normal” at intermediate distances but lighter at inter galactic distances,
emissive objects like the sun appear heavier than terrestrial objects at all distances,
but may appear “lighter” as the distance increases,
and massively emissive objects such as super novas appear much more massive than they (were).
Gravity is propagated at the speed of light (and less). & high \\ \midrule
Black Holes & 2011-11-05 & The “thermal” radiation emitted by black holes is m-figments,
which are emitted (and need to be emitted so that black holes participate in gravitation).
“Information” in the number of m, n, and p figments remains. Structures remain,
though the \textit{mnp} Model sees more rearrangements of structure than particle physics does.
Neutrinos may pass through black holes, though greater contents may reduce the odds of transit.
The neutrinos emerging would be mediators of (long distance) gravity as well. & high \\ \midrule
\raggedright Quasiparticles and Phonons & 2011-10-29 & Looking at the Wikipedia phonon article,
the suggestion that Fourier analysis proves quantum effects strikes the author as a tautology.
A finite Fourier transform on a finite field will only yield wavelengths as integral multiples
(or rational fractions at best) which when inverted will leave continuities out of the picture.
Fourier analysis is very useful for calculation and maybe for understanding "points of view" and possibly for finding valleys or peaks in potential.
The author is not comfortable with discrete analysis creating quasi particles that then interact with electrons.
Maybe better left unsaid in professional company. & medium low
\tabularnewline \midrule

\end{longtable}

unreviewed 2012-08-12

\Blindtext

\end{document}

-
I tried the adjustwidth environment from the changepage package. It seemed to affect text but not the longtable. So if I hone the global page definition to suit the longtable I want, longtable is able to fill pages reasonably. The format mostly works for the rest of the document and can be adjusted for text by changepage if necessary. Thanks to all –  Gregg Hauser Aug 13 '12 at 22:24
\newskip\LTleft, \newskip\LTright, \newskip\LTpre, and \newskip\LTpost are already executed in the package longtable. Doing it again wastes skip registers. –  Heiko Oberdiek Aug 14 '12 at 14:48
\let\LTright\LTleft: that changes the skip register of \LTright. Better: \setlength{\LTright}{\LTleft}. –  Heiko Oberdiek Aug 14 '12 at 16:02