Hi Eduardo,

There are no rigorous results to claim that selection methods are relatively insensitive to small changes in topologies, but all empirical evidence points to good robustness.
You have to get the phylogeny really wrong to really influence selection inference a lot.

If it doubt, I would simply run your analysis using several alternative topologies (e.g. you could collapse all poorly resolved internal nodes into a polytomy, assuming branch lengths are sufficiently short),
and see if that makes a difference.

Sergei