Specifically: midendian indirectly pointed out a while ago that I was incorrectly urldecoding all incoming data when I in fact should only urldecode entries. This didn't have much of an effect in most cases except that LogJam ate plus signs (they represent spaces in url strings), except that by moving the urldecoding stage to a later point, the Windows1251 data was still in urlencoded (normal ASCII) form when it hit the transcoding point so it wasn't getting converted.
Unfortunately, I'm still a bit confused how people handled other data on LiveJournal-- should fullnames on friends lists be transcoded? Does LiveJournal just pass the eight-bit data directly, then, without urlencoding? (I assume so.) But this is at least a step in the right direction, and with the feedback of someone who actually needs this functionality I think I can make it work.
Doubly unfortunately, this code is really obsolete-- anyone using LiveJournal now is probably using UTF-8 which makes all of this issue moot. But there are still other non-LiveJournal.com LiveJournal sites, and I'd like to finalize this issue before releasing LogJam 3.0.5 which fixes a few minor bugs.