Name X

Plane to plane distances have a few ways that they can be calculated.  In a perfect world, both are parallel and the distance is constant regardless of where it is measured.  In the real world, however, the planes are not parallel and we need to dictate how the measurement is calculated.  Each software package might use a different algorithm but I would bet that they take the measurement from the normal vector at the centroid of your first plane until it intersects with your second plane.  Plane to point is much more straightforward and is what your height gauge is measuring.

A 0.003" deviation is not out of the question because each feature measured has a 0.002" accuracy.  If you want to re-check it, start out with a freshly calibrated probe and try to keep the arm's joint orientation as close as possible when measuring each feature to eliminate as much variability in the encoders as possible.

Let us know how it turns out.

I often use holding fixtures to measure parts and assemblies.  These fixtures have known tooling ball locations which allow me to do a center point best fit alignment and therefore find the part. 

My problem lies with the composite position callout for a pattern of holes.  Without datums defined or used (because I use the holding fixture for alignment), The Feature Relating Tolerance Zone Framework (bottom half of composite callout) position does not calculate correctly and simply matches the PLTZF.

Any advice on a workaround would be appreciated.