Ok. I have doubled the number of simulated trials. Using the stats.norm… results are not near true values.
If I add some noise to the alpha and beta values (the ones I input to the LL function) - I’m unable to recover the true parameters whether using t or normal.