I'm currently participating in a learning seminar (a "Classics in Geometry/Topology kind of deal", on Teichmuller theory) which seems to work quite well. It is quite tightly managed, with the professor running it both picking the papers we read and giving an introductory lecture before each paper is presented. I appreciate that, since otherwise I'd have trouble knowing how to pick something interesting and understandable from the huge amount of stuff out there.
On the other hand, I can't see any overarching "goal" in the seminar we are striving towards; it's all just various interesting classical topics in the subject. IMHO, this doesn't hurt anything.