Repeatability struggles w/ short-ish part

[[Th...]]

Yes, the "drawing" is bad. Please ask questions if it's unclear and I'll try to explain. Also, this is gonna be in US Customary units (inches), so apologies to pretty much everyone. This was done on a Contura 2014, with a XT Gold head, running Calypso 6.6.28.

I'm getting variation in my test data that's higher than I'd like to see. Per the "drawing" I've attached, my origin is a circular feature, about 0.2" in diameter. The clocking feature is a short plane just under 0.5" above and 0.2" forward of the origin. Overall, the distance from the origin to the front of the part is about 3.5". Near the front there's an angled surface that I'm taking a space point on and evaluating the point distance characteristic.

Right now, with an established base alignment (i.e. I'm repeating this under Current Alignment each time) I'm getting really poor repeatability on this characteristic. Referenced to the base alignment, it's basically perfect (0.00001" range over a 10-loop repeated program, which is expected) but referenced to the B and C datum features (recalling the measured point into a new feature, not multiple physical probings), it's ranging 0.0001" at best and 0.001" at worst. The circle is cookbook standard for strategy and eval, and its X/Y values relative to the base alignment repeated in the ten-millionths range, so it seems to me that the clocking feature is what's causing my problems, but no matter what I do - cookbook standard, single-point grid, single space point with an axis distance offset - I can't get it any tighter.

Am I crazy? Are these ranges expected? I know the datum features are *really* close together, but I feel like I should be able to do better. I have line profile evaluations along the outer edge of this part that are teetering on the edge of acceptability, and when the same part can be in or out of tolerance because of the repeatability range I'm getting, I find it difficult to convince the engineers that any of my results are valid.

[[Cl...]]

Do you have a model? If so, what does that small radius just above -C- extract at?
You could you use that radius to clock (rotate by distance) instead of a point. That
might be more stable/repeatable.

[[Th...]]

There is a CAD model, which I used to program the whole thing.

The internal radius near C is 0.065", but it's probably a poor option. The entire outer profile of this part is milled, but only the flat planes are expected to be controlled. That radius doesn't interact with anything in the assembly, so there's a lot of tool deflection allowed, which results in inconsistent material condition (always plus-material). This does mean that the poor form of the radius tends to creep into the C datum feature. However, my tests with a touch-trigger grid were done far enough in from the feature edges to avoid it. Even then, the best I'm seeing is still about 0.0001" of range.

[[Cl...]]

I was thinking the OD radius above -C-, not the smaller ID below, unless it's in the same condition?

[[Th...]]

Unfortunately, it is. I seem to be limited to the planar feature only. Said plane is 0.09" deep (in Z) and about 0.15" long.