I can confirm Mike's results on phase 1.
Here follows my table which also contains the number of local maxima (which you will not find in "backward" steps):
this is true. i decided i was more interested in performance than
in knowing about local maxima.
I will try to verify phase 2 later next week.
let me offer a suggestion here. since i divided the corner configurations
by symmetry, it might be nicer if you divide the U-D edge configurations
by symmetry. (the numbers involved are the same.)
it's always nice to confirm a calculation like this using a different method.
although what i'm suggesting isn't much of a change.