Microformat parsing mechanisms that depend on documents having even minimal xml properties like well-formedness may fail when consuming non-well-formed content. Tidy may be a useful work around. In particular Brian Suda's frequently cited X2V hCard and hCalendar discovery and transformation prototypes use XSLT, and "tidy" any non-well-formed input before processing it.
Most microformats tend to be agnostic about things like exact element type used.
Developers can use tools like XPATH that assume well-formedness on well-formed content (from the web or by using tidy). Mark Pilgrim's example universal feed parser suggests that it may be possible to sanitize user html to an extent that it is suitable for later processing as xml.