Matts own documents say to read the test files for examples. He also suggests the the Perl-XML mailing list.

While it's not quite what you asked, XML::LibXML can also process XPath and it's a lot faster, and under active devlopment at the moment. It's libxml2 based so not for everyone.