I work with hydrogenases, and the amount of horizontal gene transfer is incredible. There are so many mobile elements in the environment... E coli, the humble gut bacterium, seems to have four copies of hydrogenase. Two come from the broad class of "formate hydrogen lyase"; the other two are "uptake hydrogenases". Almost certainly they come from horizontal gene transfer, because other gammaproteobacteria may or may not have some or all of these hydrogenases; and you will find uptake hydrogenases scattered through bacteria (and archaeal) phyla.
the hydrogenase that I work with is part of a mobile element; it's found in our organism in a gene island, but also straight up in the genome of another organism that was isolated 5000 kilometers away, in a totally different ecotype. both of these guys are found in the ocean, but none of the other members in the broad family that it comes from are oceanic: they come from terrestrial volcanic mats.
So obviously species-wise discrete or fuzzy taxa do not work for me; but even protein phylogenetics can be muddied by things like random gene fusions and convergent evolution. I'm not sure that some of the trees that I generate are actually taxonomic, but may be the result of selective pressure against a highly conserved scaffold.
Of course, they're still useful, even if they're bullshit... Because I don't care about phylogenetics, I care about function. If the tree that I make is an unfaithful representation of the historical record, no big deal, as long as it faitfully clusters function.