Alignin to generic feature

The big problem I've ran into when using generic features for alignment purposes is when the feature is made up of something dynamic like a variable value to define its attributes. Use logic to prevent an erroneous value from making it to your generic feature... your machine won't ask whether it should self destruct or not.

It can be done, but at Kp61dude! says it can cause problems. Generic features can do some unpredictable stuff when editing programs.

I tend to prefer using standard feature constructions or offsetting/rotating alignments by fixed values and constructing alignment planes/lines/points as needed.

In the scenario you are describing you may be able to measure a single point on Datum D while in an ABC alignment. Make sure that Snap is turned on for that single point so the measured vector is forced to be the same as the nominal vector. Then construct a cast plane from that point. It would give you a plane at the measured position of Datum D with a perfect vector relative to ABC. You can then align to that plane if desired.

Alternatively, make a new alignment that recalls the ABC alignment and translate the origin to the feature you measured as Datum D (a small plane?). Translating all three axis at the same time will put the origin at the centroid of the feature. Then, rotate the axis to the nominal angle of Datum D. From there you can construct an alignment plane that has a perfect vector relative to your ABC DRF.

I'm sure there are more ways to go about it. I just wanted to give you some ideas.