Are you multiplying the likelihood or the log-likelihood by 15? Multiplying the log-likelihood by 15 sharpens the modes (figuratively speaking) and is not equivalent to the original problem. Otherwise without looking at your code and data it is hard to say much about what is going on.