Your mike will work way better than your calipers. You are kidding yourself if you think calipers are better for anything except speed and convenience. You calipers may display in 1/2 thousandths but you are lucky to get plus or minus .001 out them accurately. Calipers are much more dependent on technique. Compare measurements of the same case. Measure it about 20 times each with calipers and mikes. By the time you have measured it 20 times with mikes you will have the technique down to probably about plus or minus .0002. With calipers you will still be getting the same old variable measurement that you think is plus or minus .0005 when it is actually much more. The mike will show you the difference. The caliper surface is also flat so you don't really have any advantage with the calipers.
Your shoulder bump varies because your technique varies and your technique is not robust enough to handle the brass variations.
This was copied from another post but it applies to any resizing operation where you find your head to datum length varying.
Sit down with about 100 to 300 cases and a Hornady or equivalent shoulder bump comparator.
Start resizing your cases with your normal process. Measure each case immediately after you size it so you have some memory of the force required. If you have small variations in the lubing technique, the speed of the resizing stroke and the dwell at the top of the stroke you will find small variations in the head to datum length (I will call it HDL).
These small variations will be about .002 max but if you are hasty and sloppy you might cause .003 variation. You can almost entirely eliminate these variations by taking a little more time.
1. Size the cases slowly. This gives the brass more time to creep (yield) to a common location.
2. At the top of the ram stroke let the ram dwell to a count of say 4 to 5 seconds.
3. Lower the ram 1/2 inch and spin your case about 120 degrees in the shell holder. Then resize slowly again and repeat the dwell again. Then lower the ram, spin the case one more time for 120 degrees and slowly size again with the dwell.
All of these motions give the brass more dwell time at the top of the stroke. The added strokes and dwell help to minimize spring back of the brass. Do this and your FL sized cases (when sized at one session) will all have the exact same shoulder bump length and your ammo will improve. The lube technique is not all of the problem. The problem is the single stroke sizing and the variations in speed. These variations combine with the lube variations to create varying head to datum lengths.
Most people cannot believe that they cause this kind of variation when resizing. Use the caliper and the bump gage to help you identify what is causing it and how to get rid of it. If you want consistent HDL you have to have a consistent resizing process and that means SLOW and REPEATED WITH DWELL.
The only micrometer I have has flat surfaces and won't work well on an angled case. The calipers I have read in 1/2 thousandths and I used the sharp edge of the blades and measured each several times while trying to slightly angle the case in the jaws to get as close as possible to perpendicular. The readings were repeatable at least to .0005". I just now remeasured the case after the temperature stabilized for a few hours and I get the same reading all the way down. A larger test batch would be better of course. I was trying to get some idea of how much I was working the brass. I don't know if I the upper portion of the die could be polished out a bit or if that's even enough to worry about.
Sometimes the shoulder bump can very up to .002" but probably in part due to the condition of the brass. Better if they are all freshly annealed. My press doesn't cam over and if I apply more pressure, it will compress a little more. I try to get everything to work from a very light stroke and then apply a bit more pressure a second time if necessary.
I'm sure the difference in size and internal volume would be beyond the scope of my ability to measure it on target but I like to try to eliminate as many variables as possible.
Thanks for the reply.