Scenario:

Select columns A and B, group together.
Select columns C and D, group together.
Select columns A through D, group together.

Select columns A through D, choose ungroup.

Rather than the group for A through D getting ungrouped, the two groups A-B and C-D get ungrouped.

This enhancement request is to either:
A. ungroup the group that best matches the selected range first (going from top to bottom as seen from the UI).
B. ungroup only the last group added that is within the selected range (which means keeping order-of-operations stack internally)

I can think of a reason why it is the way it is right now.  If you...
1. add a bunch of smaller groups
2. group those together in a larger group
3. decide you no longer need the smaller groups
4. select the range inside that larger group
5. hit ungroup.
..then you're left with just the larger group, exactly what you wanted from step 3.

With the proposed change, this would instead leave you with all of the smaller groups.  You would have to select those individually, then ungroup them.  However, a faster method for this specific use case would be to ungroup twice (removing both 1. the larger grouping and 2. the smaller grouping), then group again (re-creating the larger grouping).

A similar workaround does not exist for the current implementation; the first ungrouping removes the smaller groups before the second ungrouping removes the larger group, inherently requiring re-selection of smaller groups' ranges before grouping once more - resulting in more operations.

I may have missed something though.