> How do you prevent introducing a bias that then becomes what we "know" about how statues were painted? By introducing modern aesthetic sensibilities and present them as plausible, we then reinforce that this is how statues were painted back then, and we don't know.

This is just an argument against doing reconstructions at all. Which I am also okay with. It's not a defense of the existing reconstructions because they have the same problem. You don't want to assume additional layers. The existing reconstructions are assuming there were no additional layers. Neither are valid assumptions, but they are both possible. So present multiple possible alternatives without stating that any of them are accurate reconstructions, only that they are constructions which are consistent with the available evidence.

Surely, if one wanted to produce a "reconstruction" of the Venus deMilo, it would have arms. Even if you don't know what the arms would have looked like. And that you would not reconstruct the arms as just straight lines projecting from the stump and would make some attempt to make them realistic and aesthetically pleasing, even if the end result almost certainly does not look much like what the original arms would have looked like, exactly, it would be more like it in spirit than either the statue with stumps or with some sort of vaguely armed shaped cylindrical attachments.