similarities produced by BLAST, which imply very well
defined clusters in most of the cases. Nevertheless, it is worth
noting that our methods, regardless of the parameters used, do
not find unnecessarily large overlaps, when this is not dictated
by the data.