On the maximum likelihood method in molecular phylogenetics