When using good quality brass (Lapua, Peterson etc.) I can maintain +/- .0005 with about 85% of the cases. About 15 out of 100 will be the ones that size differently.001-.002 from the norm. With cheaper brass that % goes up to around 25-30% and out by as much as .003. Friends using coax and single stages tell me they experience the same.
If the shell plate is reasonably tightened it seems to make little difference. As long as the base of the case is allowed to sit fully on the platform of the ram as it has zero flex. When other stations have cases in them, it does slightly change how firmly the case base makes contact with the platform due to some shell plate interference. However, that difference is constant if the stations are filled with brass; if not, you'll see about .001-.0015 variation. As I stated previously, I do my sizing as a separate step to avoid that variable.
Edit: The toolhead floats as it is and I get less RO using the factory setup than a friend who tried locking his toolhead. He switched back. Runout from factory is .001 max and rare. The locked toolhead was giving .001-.002 regularly.
That seems pretty good, especially compared to what I measured. I'm +/- 0.006 across everything. About 70% of mine is +/- 0.002 with the rest outside.
I'm curious, do you have anything special with your setup? Such as, what die are you sizing with, does the die have a rubber o-ring, etc?