Yep, I don't buy the idea that the species tree is a good proxy for the ground truth for a given (short) domain alignment. But then I usually care more about branch lengths and other parameters than topology.