Common Datum Issues

I have an issue with a part I'm trying to improve. With 2020 (or I forget which exact version), they introduced being able to create a Common Datum. Well before they introduced this, we've had to basically measure feature A, measure feature B, construct a line through them, and call it Datum AB. Now that PC-DMIS has the ability to construct Common Datums, I am trying to measure Datum A, measure Datum B, and then define them both and then finally define Common Datum A-B. The issues arise when I go to dimension Position. I have tried this with a program for Total Runout, and the results are close to the same. With True Position though, it is WAY off. This part that I am working on currently has a 0.002MM True Position Dimension to Datum A-B. I'm measuring 7MM......I kid you not. I am also trying to simultaneously increase the speed on the programs, so I've tried:

• Scanning from 10MM to 300MM (No change, Size and Location of features remain same)
  
• Creating an alignment 2 different ways (Not really a drastic change, but went from 7 to 5 MM)
  
• Changing Point Density all the way from 1 to 6 (No change really)
  

What could it be? Like I said, there's no way the part can be this far out. With the old method of constructing a line between the 2 Datums and calling that the "Common Datum" checking good parts, I wonder if the math is somehow messed up with this Common Datum construction in PC-DMIS?

Thanks