I have had luck dealing with the tree comparison problem by treating each edge
as separating two groups of taxa. Two edges in two trees which separat the same
taxon clusters are the same, and two trees which share all the same edges, defined
as such, are identical. Basically pretty similar to Joe's consense algorithm.
Seems to work fast enough. I don't have it written to read in trees, but it
probably could be patched into the code from consense fairly easily.