Prompted in part by the question Is there an "underscript" command?, I started wondering if the \underset (from amsmath) and \underaccent (from accents) commands work equally when using XeLaTeX instead of pdflatex. The following code and associated output show that this is not the case for the XITS Math package:
\documentclass{standalone}
\usepackage{amsmath,accents}
\usepackage[no-math]{fontspec}
\usepackage{unicode-math}
\setmathfont[version=cambria]{Cambria Math}
\setmathfont[version=lm]{Latin Modern Math}
\setmathfont[version=asana]{Asana Math}
\setmathfont[version=xits]{XITS Math}
\newcommand\undersetunderaccent{$\underset{x}{A} \ \underaccent{x}{A}$}
\begin{document}
\Huge
\noindent
\mathversion{lm}Latin Modern Math: \undersetunderaccent\\
\mathversion{cambria} Cambria Math: \undersetunderaccent\\
\mathversion{asana} Asana Math: \undersetunderaccent\\
\mathversion{xits} XITS Math: \undersetunderaccent
\end{document}

Clearly, the output of the two commands is rather different [!] when "XITS Math" is the math font. Does the XITS Math package have to be loaded with special options to suppress this behavior, or should I plan on sending a bug report to the maintainers of the XITS Math font and/or the maintainers of the unicode-math package?
Addenda: (i) I can generate the output above using both TeXLive 2011 and MiKTeX 2.9 (both updated yesterday); the OS is Windows Vista; all math fonts on my system are the latest ones available. (ii) The problems with "XITS Math" do not arise if I run the example code above under luatlatex instead of under xelatex.
lualatexthough. – Torbjørn T. Nov 16 '11 at 19:25