A few things to consider...
A simple thing to try would be what D. Tubb describes in one of his books... skip a station when inserting cases i.e. put a case in station #1, size it, rotate to station #2 for powder but *do not* put a new case in station #1, rotate it to station #3 (which has no die in it) and *now* put a case in station #1 for sizing, then rotate again so you have a case @ #2 (powder die) and at #4 (seater die) but nothing @ #1 and #3. The idea is to isolate the step that requires a lot of force and robs what little sensitivity the press has (sizing) from the more delicate steps (powder charging and bullet seating) that benefit from more feel and don't put much stress on the shell plate. It slows the overall rate down some, but may help with more consistent results.
Another thing that Tubb mentioned (I hesitate to say 'pioneered' as someone else may have came up with the idea, but I and many others first saw it in his book) was to 'float' the dies in the tool head by use of a locator pin btwn the tool head and the lock ring - the dies aren't screwed down tight to the tool head, but instead are allowed to float a bit in the 'slop' of the coarse 7/8-14tpi thread to find their own center. Modern machining is good and all, but the odds of all four holes having coplanar centers is minimal at best. John Whidden (a top-ranked NRA Prone shooter who also runs Whidden Gun Works) offers tool heads all set up with the necessary holes drilled, roll pins, modified Forster die lock rings, and a powder die w/ funnel for those of us who want to use a 550/650 but still weigh our charges first.
Finally... UniqueTek markets a couple goodies that help out as well. One is a replacement star wheel for the shell plate with a bearing plate installed - the idea being you can snug the die plate down a bit more but still be able to turn it. Not sure how much in terms of absolute measurements this actually helps, but in concept it makes sense. The other is their toolhead clamping kit: this is a set of drill bit, thread tap, and heli-coil insertion tool to take up the 'slop' in the toolhead-to-frame fit. The modifications are made to the aluminum toolhead, not to the press frame, and the kit comes with enough helicoils to do multiple tool heads (4-5, IIRC).
So... here's my story: I decided to try the Whidden tool head for my 6x47L tactical rifle. Everything was working fine, except I was getting between -0.004 to +0.006 variation in seating depth. If I was jumping the bullets 40-50 thou that'd be one thing, but I was seating +0.010 into the lands (even with magazine feed) so I wasn't too happy about that. I talked with John via email a bit, and he suggested getting the toolhead clamping kit. I'd previously considered and discarded the idea under the premise of if floating the dies was good, then floating the dies and toolhead must be better. On John's recommendation, I ordered the clamping kit, and set up a tool head with it. With all the dies back in the tool head, I now saw +/- 0.001 seating depth variation over 20 rds, with one exception (0.005) which was opposite a case that sized hard (wasn't using the skip-one strategy mentioned above). So from what I can see, between using the skip-one method, along with floated dies and a clamped tool head, you can reduce the variation considerably.
FWIW, I believe at some point this spring Whidden is planning on offering CNC machined tool heads with a tighter fit, and perhaps selling them through UniqueTek with the helicoils pre-installed. Win-win as far as I'm concerned
YMMV,
Monte