FYI, assertAlmostEqual.m (I erroneously added an 's' to the end in my earlier comment) is not in the 'xunit' directory in the zipfile. You can fetch it manually from the menu (right), but it uses 'mtest.utils.isAlmostEqual()'. If you manually change this to 'xunit.utils.isAlmostEqual()', it then works OK.
I think I've figured out why subfunction tests don't work (at least, not in R2011a). There are a number of problems. When you put in the call to the 'initTestSuite' script, it seems that the (undocumented?) '-subfun' option to 'which', which it uses to discover the subfunction names, just doesn't work, and always returns an empty array. Then, even if you can hack it to get the names of the subfunctions reliably, which I did by hacking Robert Bemis's 'SubFuns', it seems that you're in a "Catch 22": the initTestSuite script has its own variables, of course, but in the context of a function with subfunctions, adding new variables is not allowed, apparently. OTOH if you convert initTestSuite to a function, it is not then running in the context of the function with subfunctions, and the subfunctions aren't then visible, and the 'str2func' on their names (even though you now know them) fails.
Hope this helps someone save the day I just wasted.