I really appreciate the quick and detailed response!
The Rican et al 2013 article "Biogeography of the Mesoamerican Cichlidae" didn't offer any specific analysis about grammodes, since it was so broad and comprehensive - but it did show grammodes grouped in a cladogram with Paraneetroplus, Herichthys, and Theraps (the "herichthyine cichlids")- but diverging 20.1 million years ago in Fig. 2! The same results showed the "amphilophine cichlids" including Parachromis, Petenia etc. diverging from the herichtyine cichlids 31.8 million years ago, but the article notes that "the basal nodes of the amphilophines lack significant support." Simply stated, I'm skeptical about the ability of genetic analysis to accurately sort out a cladogram going back that far! Part of my suspicion is based on the tendency of species to exchange genetic material via hybridization as well as via other means (such as viral exchange). Another part of my suspicion is that these authors were primarily motivated by finding evidence in favor of their biogeography model (which admittedly is probably a pretty good model); I'm afraid poor little grammodes was a peripheral topic for them. All I'm really saying is that it would be interesting if the "basal nodes" of parachromis and grammodes could be explored further.... Maybe grammodes was a result of some inter-genus hybrid, after some parachromis "basal stock" crossed some shallow seas and the herichtyine cichlids started looking pretty good to them....
For example, one very legitimate hypothesis for the origins of the coast redwoods is an ancient hybridization event between Metasequoia and Sequoiadendron (although now I'm talking about apples instead of oranges...)
Absolutely there could have been convergent evolution leading to an efficient herichtyine piscivore, however regarding the visual differentiations, please pardon me but those don't seem like very strong arguments. As an aside, I've found sexing very young dovii not particularly difficult!