3DDave, I am trying to visualize what you are saying. I attached a picture of the part I am working in case you wanted to reference it. The pins in the drill plate that you are talking about would be the same size as MMB, so .091 in my case. As the .096 hole grew to say .101, it can now move and rotate about those same .091 MMB pins.
My thinking is that regardless of what the hole diameter is, it is still able to move the center of the datum holes by half of what the positional tolerance / boundary condition is. So at .096 with location tolerance of .005, the center is able to shift by .0025 (and therefore the two holes are able to shift the same amount). Then at .101 with locational tolerance of .010, the center is able to move .005 (.010 / 2, half one way half the other way)(and the two holes are able to shift this same amount).
So it seems like then the two holes (I guess it is more clear to not say the two holes, but the positional tolerance of the two holes), if referenced off of these 4 holes, if you just think about it in an x or y direction, would also be able to move the .0025 or .005 depending on the size of the .096 nominal datum holes. So worst case for both features, you would add the possible movement of the datums at LMC with the possible movement of the two holes correct?
It seems like you are saying that you can add the tolerances but you say you cannot.