This paper seems very reasonable, but there's a potential problem with HGT genes and rapid divergence. Recently acquired genes do indeed evolve more rapidly and under weaker selection (e.g., Daubin & Ochman, Genome Research 2004; Hao and Golding, Genome Research 2006). These genes may also be phage-associated, which could increase divergence rates regardless of the strength of selection. These recently acquired genes are precisely the ones that would appear in one genome and be absent from relatives, and hence potentially as false positives. It would also be interesting to know if these questionable model hits are ORFans.