Unfortunately, this stuff is buried pretty deep in the data.
Early on, we noticed that Whipple had more window than stock and Roush so my buddy, who is a hex code guy, compared the bin files and identified a difference regarding variable pfi window. We switched it on in a Roush tune and the window expanded by about 2 ms. A transition appears to take place when you near the window limit, and the window expands in a Whipple OS and we just figured out how to change that in stock and Roush OS. 2 ms at 7500 rpm is worth about 12% in the duty cycle. Of course we tried a few dozen other things before that and none of them worked.
The other change wasn’t as easy to find. As I understand it, the logic is constantly calculating how much fuel to inject on a current-cycle basis. Anything injected after the air stops entering the cylinder doesn’t count towards the current cycle, so it cuts off the pfi window there. However, we know physically that fuel injected after that point in the cycle counts toward the next cycle, so I extended the window, again with the help of my hex guy. The additional fuel is just one cycle delayed which I believe is adequate in these cases.
Obviously, this isn't something you can just do using standard HPT definitions, but it didn't require any sort of patch to make it work either.