So if I were to read the time off the game clock with the pause trick active, how far off would I be? If I know this, I can post some actual results without the need for a "fixed" version.
I still need to find out whether the clock display in the real game reflects the internal clock variables on the frame or on the previous frame. If we take the current Lemmix model, where the time you see in a frame is exactly the values in the internal clock variables, and ignore the thing about ESC and/or high-perf, then right now, I think Lemmix is 3 frames too early (2 from not being able to pause on first 2 frames in real game, and 1 from the entrance opening one frame early in Lemmix). I think.
For something as simple as Fun 1 and Fun 6, if you really want some results now that are guaranteed replicable in the real game, you can use my "
vgalslow.exe" slowed-down version of DOS Lemmings. It'll be a bit of a pain to use due to the slowness, but should be tolerable for short Fun levels.
Another thing to keep in mind is that if your solution is optimal in terms of number of frames, then it is an optimal solution regardless of how the game clock is treated, just that your reported result in terms of the game clock might be inaccurate.