Question regarding whether the BEST model can be improved

Questions
2

Essentially, I’m asking if it is normal that the Normal distribution does better at hypothesis testing for T-distributed data than the T-distribution itself.