No good ones.

It isn't mixed NEC1 and NEC2. It isn't the minor difference in timing.

I think the Pronto Hex we're using describes the same IR signal as the original remote sends. But you test and the result is different.

There's always the chance that the NG firmware has another bug that I'm either forgeting or never understood. But the only way to investigate that (and look for work arounds) is to learn the signal from the NG Pronto into some other (preferably more trusted) IR learning device with PC download ability. I expect you don't have one of those.

The method you started with (simulate many very short presses rather than one long press) may be the easiest answer, even though we now see it is quite different from what the original remote sends.