A friend and I looked at AA% a while back and thought that BS uses the average attentuation %. This was purely based on swapping from high attenuators to lower ones and watching the stats change.
I just created a basic small-batch test, swapping 7# of 2-row (Briess; PPG of 1.037) for 7# of roasted barley (generic; PPG of 1.025), at 75% EE%.
For the OG/FG numbers, I got 1.053/1.013 (pale) and 1.036/1.009 (roasted), so it would appear to be working on the OG. 32% less PPG led to 32% less OG.
I don't know about pure lactose, but I 'believe' that the diminished fermentability of roasted is reflected already in the 32% lower Potential Points per Gallon. Roasted barley, the crystals, etc., are "steep only" grains, so the sugars already converted by the malting process, although lower, should still ferment out at the same AA%, here being 75%. That seems correct here, going from 036 to 009.
The two grains you mentioned are PPG of 1.036 and 1.034, much closer, and you added an extra 1/4# of the 80L, effectively compensating for the slightly lower PPG, thereby equalizing the total OG and FG from each grain, as your numbers indicate.
The crystals, roasted, etc., are slightly less fermentable and that's reflected in their PPG. I think where we start talking about significant "unfermentability" is when the proportion or raw amount of crystal grains gets large in relation to the base grains, as in your SEB, with 2.5# crystal in 8# total, or 31%.
Hope that helps.