If you open San Francisco Crime.jmp and run out Day of the Week across the top and Category down the side, you can see this interaction perfectly.
When you do this, go ahead and paste in the above stat testing schemes. What you will notice is that when you apply the second scheme (A/B/C,A/D,A/E,B/D,B/E), comparisons are similar that were previously flagged as significant no longer are.
A/B/C/D/E -> Vandalism -> Column A stats against C,D,e (UPPER 90%, LOWER 80%)
A/B/C,A/D,A/E,B/D,B/E -> Vandalism -> Column A now only stats against e (LOWER 80%) even though A is still being compared against both C and D