I have to agree that you are not seeing the same addition of the two groups that I did.
I was using a D200 when I ran my tests. I now have a D3. I should try it again.
My previous understanding of the system timing was that each group was analyzed, and a power calculation was immediately sent out to that group, in which case it could not possibly know about any of the other groups, and therefore could not take their contributions to the image into account.
I was looking at your timing sequences in your part 2, but I don't see any analysis of the 'remote flash output amount' when there are two groups in use. Did you study this sequence to understand the format? I would assume it is an address followed by a power setting for each group.