"Default" Math Vs "Legacy"?

If the Customer gives you Datum bonus, and your Datum bonus is enough to allow your Datum to shift enough to give you a zero deviation on your feature, then that's what you get. If you or your customer doesn't like it, then remove the Datum bonus.
Judging by this post, I'm fairly positive your customer doesn't trust you, not your software

SingularitY is correct. The whole point of adding material modifiers to datums is to allow datum shift and, if there is enough datum shift available, then the position will fully optimize to achieve a zero result.​

I think more correctly, the topic I'm trying to bring to discussion is:

"New Math" performs a best fit calculation, if you are giving direction to a set-up machinist, I would consider using "legacy".

What are your feelings?

To​ Cumofogabriel 's point, yes, absolutely. If you are looking to provide information for process adjustment then using legacy in the way you have done will negate any datum shift and let the machinist know what corrections may be required in order to improve results. But, as I said, it is not verifying the callout in accordance with ASME Y14.5 as the designer intended.
​
The reason for my taking offence at your initial post was that you didn't explain this might be what you were attempting to do, you just launched straight in with what I perceived to be an attack on the geometric tolerance command. Since I've personally invested the last four years helping to develop, test release and support the "new math", I tend to take those kind of comments a little personally.

The way PC-DMIS expresses datum shifts by reducing measured values is stupid, IMO. Would have been much better to add to the tolerance, rather than subtracting from the measured values.

In those rare occasions, where PC-DMIS can calculate in tolerance vs out of tolerance conditions correctly for characteristics with datum modifiers (no software can out of the box), one can report twice, one with datum shifts active (to determine red vs green), and once with datum modifiers inactive to provide real data.

The way PC-DMIS expresses datum shifts by reducing measured values is stupid, IMO. Would have been much better to add to the tolerance, rather than subtracting from the measured values.

In those rare occasions, where PC-DMIS can calculate in tolerance vs out of tolerance conditions correctly for characteristics with datum modifiers (no software can out of the box), one can report twice, one with datum shifts active (to determine red vs green), and once with datum modifiers inactive to provide real data.