I noticed the same. Those files should be easy to recognize. Most of them have a name matching "MANIFEST*.MF" and all of them start with the text "Manifest-Version:". For reference: the format specification is available at as part of the JDK.

These seem to be duplicates too: http://ohloh.net/projects/3374 (which is empty) and http://ohloh.net/projects/3375 (which seems proper).

I recently cleaned up the repository of http://ohloh.net/projects/740 using svndumpfilter. It contained a lot of support material for building the documentation, which resulted in a mostly-Java ... [More] project to have 69% XML. To my understanding an incremental update of the stats will not realize that the XML files are gone (as they were not removed using "svn remove"), so the stats will remain with 69% XML. Please change the next scheduled download to get the full repository instead of just the changes in order to rebuild the stats from scratch. Thanks! [Less]