Phew..... got her back to 0.005". I tried all three possible chuck orientations and the best was what I thought was the original. Still I had to tap her over a good 0.02". Need to mark this orientation permanently now. So, I guess this would mean I have to turn 0.002" off the mounting plate boss in order to attain 0.003". Does that sound correct? I don't feel confident enough in my machining skills to actually attempt it.
I’ll give an example. Let’s say you have a dial test indicator (DTI) resting on the OD top of a perfectly accurate hardened dowel pin held in the chuck jaws. It measures +0.001” (high) and -0.003” (low). We calculate runout by (low – high) / 2. Runout = (-0.003 – 0.001) / 2 = (-0.004)/2 = -0.002”. This means the dowel axis is 0.002” lower than the spindle axis. Remember to obey the corresponding low & high sequence order in the brackets and also the +/- sign, then it always works regardless of DTI readings. I've attached a sample of different reading combinations. It doesn't matter if you null the needle to zero at a low or high position as long as you follow the procedure.
Now we have established runout and want to correct the chuck so the dowel axis is 0.000” meaning coincident with the spindle axis. This presumes the chuck recess is completely flush tight against the back plate lip in the offending direction, i.e. you can’t alter it anymore to improve the runout. You need to remove the same 0.002” on the offending side of the back plate lip in order to displace the chuck that amount. But that is equivalent to a radius which means the back plate lip diameter must be reduced by 2 times that: 2 * 0.002 = 0.004”.
Notice we are talking about runout of the dowel reference part held the chuck jaws, not the chuck body or any other unrelated feature. However, if the chuck jaws grip the dowel inconsistently or non-repeatedly, then all bets are off. This problem is common in 3 jaw / scroll chucks, more-so older, worn chucks. Count yourself lucky if you are within 0.001” runout. Before you do any back plate machining or jaw grinding, check this first. Grip the dowel, measure runout, mark the amount and direction with a felt pen on the chuck. Remove the dowel, unwind the scroll & repeat a few times. Ideally same tightness & across all key slots. Also chuck a larger diameter dowel to see if readings vary. If the readings are consistent amount and direction, its probably chuck/spindle alignment. If the readings are dancing around, then its probably jaw related (assuming spindle is in good shape). If the chuck scroll is in good shape, maybe you can buy a set of new jaws & be back in business.
If jaws are unobtanium or the chuck is a direct screw mount to spindle with no intermediary backplate (lip), then jaw grinding is probably the next step. Some guys have no problems taking this on, but do some research as to equipment involved & how to properly pre-load the jaws.