I've looked at Plink a few times recently, and was puzzled by the lack of LiveJournal.com influence on the links. It seems that Plink depends on the mbox hash to make the links, and although we publish rdfs:seeAlso elements pointing at a file actually containing said element, it doesn't look like plink is able to follow them to make the links. This makes sense, really, since FOAF uses anonymous RDF entities and leaves it up to software to make the match, and any given RDF document can contain any number of entities, so we aren't giving any formal link at all between the entity which has the name “mart” in my friends' FOAF data and the entity named “mart” in my own FOAF data. All Plink knows is that there are two entities nicknamed “mart” and one of them is a friend of, for example, “brad”.
If we were to add mbox_sha1sum elements to each “knows” element then the links on Plink could perhaps get more interesting. I'm not sure that this is what's causing Plink difficulty, but given that it seems to do all of its matching on the basis of email address (or hash thereof) I think there's a pretty good chance that this is it.
The lack of “real names” in LJ's FOAF data makes it hard to search for people on Plink, too, but there's not much we can do about that since our real name field really means “a load of irrelevant junk” in practice. At least people will be able to start from Brad Fitzpatrick (who was given a name through someone else's FOAF data) and browse the network from there! (It's also possible to “search” by FOAF data URI and find LJ people's nicknames, but they don't make that obvious)
Any reason why we shouldn't/can't add the mbox hashes to the “knows” elements?