I am trying to understand the evolution of different bacterial genes. For each of the genes, I have sequence data from different isolates and want to construct a phylogenetic tree that is reliable (or at least the least unreliable). There seem to be very different opinions about which method is best to use, some people stating the maximum parsimony is THE method for single gene trees, especially with highly similar sequences, others stating that maximum parsimony may produce avoidable errors due to lack of parameters and is computationally expensive, recommending maximum likelyhood or bayesian inference instead. Which method would you use considering datasets of highly similar nucleotide sequences with alignments from 10 to 1000 sequences, and why? I would be happy about any input!