That "accuracy" would get carried thru every connected constraint - multiplied by distance and potentially screw up designs. Just put a plane thru the center of both holes (by way of axis on the centers). Then constrain one end hole to hole and then plane to plane to stop rotation.
It would not do what you say.
First, we are talking about cases where there is no mismatch visible in the 6 places Alibre allows you to see. The error is down in the noise, probably due to Alibre having a rounding error.
Remember,
location would be aimed at "precise", but would
tolerate an error if the part could not be moved into compliance. So all parts that fit will be right down central.
How would that differ in the slightest from doing what you suggest (which is a workaround I have used for years)? The part is still the same, the location is still the same, the same errors occur if you base other parts off that "misaligned" hole.
The difference is in not having to fiddle-fart around constructing a bunch of axes and planes to get it to locate. That is a pain when there are multiple cases.
And permit me please, to remind you that real-world parts have tolerances, and products get built. If you need the part to 0.01mm, put your tolerance to 6 places. That will never cause a real world issue in a reasonable design, unless your shop works to microinches or nanometers.
This differs from the present "limits", because it would not be "dynamic". Limits are dynamic.
And this entire problem derives from a fundamental issue in Alibre, which is that "fit" is judged by the program as being accurate to many more decimal places than Alibre will allow you to see or input. You have no control over it.
Calculate a diagonal on a square frame some time. If the diagonal true distance has more than 6 decimal places, Alibre will probably not allow the alignment.
Here, play with this one.... the dimensions are identical..... as far as YOU know.