That would correspond exactly to the formula I gave for the DOS version, after accounting for the lower resolution on the DOS version compared with Windows.
No, sorry my fault. Indeed I use "8 + (99 - releaseRate)" since I update at the double refresh rate to compensate for the doubled resolution (don't ask). To avoid confusion, I wanted to reverse this value for discussion to avoid confusion, but only divided the first part by two.
So indeed my "8 + (99 - releaseRate)" at doubled frequency means the same as "4+ (99-releaseRate)/2" at normal frequency.
That isn't necessarily messed up, I never bothered to measure the game time of Amiga or DOS Lemmings against real time, so they might differ as well. In fact, it's almost guaranteed that at least one of them does not follow real time, because on levels with tight time limits, for example "Just a Minute (Part 2)", some solutions that work in DOS version times out on the Amiga.
Interesting. Yet, if a game tells me that I have 5 minutes left, if I can nearly use 7 minutes, I would call this a messed up timing.
If my theory is right and the PAL Amiga version used 60ms per period, it would have most probably used 17 ticks for one second. So one minute would be 61.2 seconds. With 66ms per second and the same number of ticks, it's already 67.32 seconds.
Assuming, that only the time resolution was changed, Amiga PAL and NTSC (or DOS/WIN) versions would differ by about 6 seconds per minute. However the Windows version is off by 20 seconds or so.
It's generally accepted in gaming that game time does not correspond exactly to real time. Often people don't even realize it during play.
I don't really like that idea to be honest. Time will be exact in my port. Probably I will give the level time in seconds, so levels can be "fine tuned" to the second, without messing around with the length of a second.
I will advise to use a speed that matches the Amiga version. The main rationale being that for the level "Just a Minute (Part 2)", there is good indication that the easier solutions that work in DOS version but times out on the Amiga likely indicates that such easier solutions are not intended to work, that such solutions should in fact time out like in the Amiga. Thus if your game clock runs slower, it would be violating the spirit of the level (unless you choose to modify the level to compensate).
On the other hand, if your game clock ends up faster than the Amiga version, then you should test out that level to make sure it remains doable.
I'd guess that the clocks in all versions run (more or less) slower than real time. Mine will run normally (no FF, no Superlemming) at real time, so it will probably be faster, not slower. However the difference to the Amiga version should be minimal (1.2 seconds per minute, if my assumptions are correct).
If people will complain about certain levels in the beta testing phase (hint ;D), I will patch the level time to be a little longer.