If the model wasn't optimized with a parameter = 3.08 (that is, the optimization figured 3.08 increased the objective function), why would it be better than parameter = 3.1 ?

(This applies to functions of parameters too)