XeLaTeX font choice for separable ligatures in PDF, redux

Question

I've read previous question/answers on this topic (e.g. this), but am still in the dark. The problem I'm trying to solve: Which font combination for XeLaTeX will at the same time provide a legible and attractive visual aspect of the final document and the ability to copy/paste out of and search inside the PDF file. The issue is the separation of ligatures.

Here's a small example of my .TEX file:

\documentclass[12pt,letter]{article}
\usepackage[margin=1in]{geometry}
\usepackage[sf,bf]{titlesec}
\usepackage{parskip}

% Font settings
\usepackage{fontspec}
\defaultfontfeatures{Ligatures=TeX}
\setmainfont{Minion Pro}
\setsansfont{Myriad Pro}

%% The document
\begin{document}

\section{Introduction/Question iffy fig there office baffle}
This reports continues from the initial report. It contains answers to questions, follow-up work etc  iffy fig flat office baffle

\subsection{Slope and aspect map generation iffy fig there office baffle}
There were some pitfalls discovered when generating slope gradient and aspect maps from the ASTER DEM using the GDAL command line utilities. Ähnlich Schoß. μ δικαιώματα in Greek. $\mu$ in math.

\textbf { Questions on Slope/Aspect Maps iffy fig there office baffle}

Should the cut-off for ``flat" be selected in a more scientifically founded way? 

The three sub-regions are not very homogeneous: there is "noise", that is, isolated pixels or small structures within one region (say, ``north") that are classified as a different region (say, ``south"). Should these be smoothed (via a median filter, for example, or a morphological filter)? 
\end{document}

I have tried this with the combinations

• Minion Pro/Myriad Pro
• Linux Libertine O/Linux Biolinum O
• Cambria/Calibri

Any of the three would be acceptable for my purposes, though I like the Minion/Myriad combination most. But each of them fails to separate different sets of ligatures. Cambria is overall worst. Libertine/Biolinum is good for he fl/fi/ffl/ffi ligatures, but the (beautiful) capital Q in Libertine, and capitalized Th are an issue. Below is a screenshot of the result of copy/pasting from the PDF output.

Are there any settings in fontspec or \defaultfontfeatures that I could change?

Answer 1

I think you have two options:

• Switch from XeLaTeX to LuaLaTeX. When I run your MWE on my system (MacOSX 10.9.2, MacTeX2013, Acrobat pdf viewer, TeXworks editor), I get the same problems as you do when I compile it under XeLaTeX, with all three font pairs you mention. However, no problems occur if I compile the MWE under LuaLaTeX using either Minion/Myriad or Libertine/Biolinum. (Quite a few problems remain with Cambria, though, even under LuaLaTeX.) LuaLaTeX can only handle oldstyle numbers, if enabled, with Minion, Myriad, Libertine, and Biolinum: if you copy-and-paste oldstyle numerals from a pdf file into a plain-text file, they should show up correctly (probably as lining-style numerals) in the plain-text file.

• If you must stay with XeLaTeX, use a font family that's "well behaved," by which I mean that it'll let you cut and paste from compiled pdf document to a plain-text file under XeLaTeX without generating weird glyphs where various deligated glyph pairs and triples should be. I'm afraid I'm not aware of too many such font families. :-(

  One such serif font face is EB Garamond, two others are Dante MT Std and Sabon Next LT Pro. (Note: Dante and Sabon Next are not free-of-charge.) Dante and Sabon Next feature medium and bold weights, in both upright and italic shape. A possible downside of Dante and Sabon Next is that they provide exceedingly few Greek text glyphs (only one or two, I think). As @Sverre has pointed out in a comment, Junicode is also "well-behaved". :-)

  A sans-serif font that meshes reasonably well with EB Garamond, Dante MT Std, and Sabon Next LT Pro is Palatino Sans -- again, not a free-of-charge font.