In my early days, 1979 , with no comparator or cartridge gauge, the Dillon shell plate gave me .014" shoulder set back and case separations in my M16A1 carbine.
That is interesting. I didn't think about there being that much variance in the thickness of the shell plate and I bet that's another factor in the variance I get, not just slop in the machine itself. I also adjusted the bullet seating die down to the same amount of touch and I size all my brass at once without rotating the shell plate. Doing this I get decent tolerance.
But now thinking about all that maybe I could adjust my die up to bump the shoulder less, and not rotate the shell plate and get acceptable variance (no more than +-.003)