You'd need to run anything from MySpace through HTML Tidy first before hitting it with XPath - and even then, you're asking for pain.
I feel dirty, but I have built MySpace spiders using Python and BeautifulSoup before. Worked somewhere that was doing an ad campaign based on MySpace, and they wanted social network metrics.
IN any case... MySpace is pain.
But in the general case, good luck on the XPath database. Not sure, but this old code might be useful:
http://decafbad.com/trac/wiki/XSLScraper
