At the moment it works for lit, html, rtf, txt, & lrf. Will be adding pdb in the next release. I wasn't concerned about epub because most of the bad epubs floating around were sourced from one of those formats.
However the launch of the ipad & ibooks seems to have incented the darknet users enough to release torrents of crappy epubs lately (converted from the source formats above). Problem is that the bad epubs are already split into 260-300kb chunks at arbitrary points, so unless I also look into recombining them into a single flow they'll still be fairly crappy even after preprocessing.