Same likelihoods, different convergence points
There are cases in which the H1 likelihoods , under similar settings, are the same between fastcodeml and codeml while the estimated parameters are not the same ( more specifically, the w2 values are meanigfully different ).