My guess is that the issue has more to do with the form of your likelihood that the value of sigma. The likelihood suggests that your observations are normally distributed around the empirical mean. But your data doesn’t appear to be normally distributed (to me). Have you considered other possibilities?