In the following example I played around with the sort order of nomencl by using the <prefix> command:
\nomencl[<prefix>]{<symbol>}{<description>}
However, it required to transform all the nomenclatures to the same type (string in this case). That is, using the prefix for each item (in text/non-math) so that they are sorted in the same group.
\documentclass{article}
\usepackage{nomencl}% http://ctan.org/pkg/nomencl
\makenomenclature
\begin{document}
\section*{Main equations}
\begin{equation}
a=\frac{N}{A}
\end{equation}%
\nomenclature[a]{$a$}{The number of angels per unit area}%
\nomenclature[N]{$N$}{The number of angels per needle point}%
\nomenclature[A]{$A$}{The area of the needle point}%
The equation $\sigma = m a$%
\nomenclature[s]{$\sigma$}{The total mass of angels per unit area}%
\nomenclature[m]{$m$}{The mass of one angel} follows easily.
Let's try with another $a$... %
\nomenclature[a1]{}{The area of the needle point}%
\nomenclature[a2]{}{The distance to the moon}%
\nomenclature[a3]{}{The number of needles in a haystack}%
\nomenclature[a4]{}{The value of $pi$}%
\printnomenclature
\end{document}
In the above example, all items under the a nomenclature that should be printed without a, has been given a prefix a? where ? is a numeric to force some sort order.
Text-only nomenclatures can be used in the traditional form without the prefix <prefix>, like (say) \nomenclature{C}{This is a C, see.}; included in the above MWE. You'll notice that the \sigma entry has been sorted based on the text prefix s. This can be changed if needed.
Since there is no connection between the items in the List of Nomenclatures and the grouped sorting has been manually altered (via specifying a1, a2, ... for items using the same symbol as a), there is no way to easily accommodate the following request automatically:
Manually, however, it is no problem. Just add the respective dashed (-- or en-dash) before each item. This should not be a major drawback, since the nomenclature items are formatted and inserted manually in the document anyway:
\documentclass{article}
\usepackage{nomencl}% http://ctan.org/pkg/nomencl
\makenomenclature
\begin{document}
\section*{Main equations}
\begin{equation}
a=\frac{N}{A}
\end{equation}%
\nomenclature[a]{$a$}{--\ The number of angels per unit area}%
\nomenclature[N]{$N$}{The number of angels per needle point}%
\nomenclature[A]{$A$}{The area of the needle point}%
The equation $\sigma = m a$%
\nomenclature[s]{$\sigma$}{The total mass of angels per unit area}%
\nomenclature[m]{$m$}{The mass of one angel} follows easily.
Let's try with another $a$... %
\nomenclature[a1]{}{--\ The area of the needle point}%
\nomenclature[a2]{}{--\ The distance to the moon}%
\nomenclature[a3]{}{--\ The number of needles in a haystack}%
\nomenclature[a4]{}{--\ The value of $pi$}%
\printnomenclature
\end{document}
