Rico Morgenstern comments on the subject.
Before I knew of the misidentified sequences I also thought that some hybridization event may have led to the inconsistent position of T. salvini in molecular phylogenies. However, if we disregard the results based on those wrong seuqences, we have to acknowledge that the sister group relationship between Trichromis and Thorichthys is unambiguously supported by molecular phylogenetic analyses across different datasets and methods, including the recently developed Next Generation Sequencing approaches. There is not the least indication fof a hybrid origin any more. Therefore we can safely say that, according to our current state of knoledge, Trichromis is indeed most closely related to Thorichthys. On the other hand, a close relationship to Amphilophus trimaculatus or any other amphilophine (ad some stage I had also Parachromis in mind) must be regarded as untenable.
This is a nice example to illustrate that similarity (which in this case is anyway rather superficial) is not the same as (phylogenetic) relationship. Closely related species are often similar to each other, but similar species are not necessarily closely interrelated. Similarity can have different causes such as convergence/parallelism, or the retention of ancestral character states. After all, herichthyines and amphilophines share a common acestry, and it could well be that their common ancestor was a generalized carnivore/piscivore like T. salvini and A. trimaculatus. The same 'type' is represented in other linages as well (e.g. Mayaheros or the Parachromis friedrichsthalii group among Amphilophines, Chiapaheros among Herichthyines), and we should not forget Nandopsis or the Kronoheros-Heroina-Caquetaia clade, which stand in some way basal (exact position needs further investgation, though) to the Herichthyins and Amphilophines.