I also think that simulation models, if well designed can give a pretty good insight into what will happen in real life and will probably give pretty good results in majority of rifles.
Pretty good - well probably not sub half MOA, but well within minute of deer at normal hunting ranges. And majority of rifles - off the shelf factory rifles in good order made in the middle of production runs so well within tolerances. Your shagged put barrel with pitted bore and long lead will be at one end of the bell curve, your precision custom at the other.
To get a rifle to achieve its full potential may well need quite a bit of tuning to really shoot well.
And then of course there is the argument over whether all of this is statistically valid and whether its worth the effort? Fundamentally this is down to the individual and own particular desires, circumstances and needs.
Using the Lee Loader with a hammer and yellow plastic measuring cup will give perfectly adequate ammo for most hunting needs, so to can taking the optimum load from a simulation model. You can certainly put meat on the table. But can you confidently shoot that buck of a lifetime at 400m, or be competitive in F Class world championships - perhaps, but probably not.
As for the statistic arguments - I will leave that to those who teach the subject to explain.
