Diameter deviation is important, and at the same time less important.
Starting from a non-round stone, and trying to get a pointed, centralized culet, and maintaining an even girdle, you cannot have the same angle on all main facets.
Therefore, the more round a stone, the easier it is to achieve optimal symmetry, and thus a nice H&A.
You have some arguments that some amount of asymmetry might add to the light performance, but I think that even if one has a perfectly symmetrical stone (according to the current measuring methods), there will still be some asymmetry.
An intelligent cutter would make sure to have a very small diameter deviation, when going for the best possible cut. On the other hand, an enormously talented can still make a great cut out of a slightly unround stone.
As for Garry's remark of 10 to 20% having an unnoticed diameter deviation because of the original triangular shape of the rough, it might theoretically be possible, but if it happens, I am sure that it is rather exceptional, and definitely not around 10%.
I will keep my eyes open in order to find an example. If so, Garry, I am sure that you would buy it from me for educational purposes?