My guess would be that the game code itself is setting the playback time slots for the audio. In order to get the proper timing set for the Japanese audio, it would need to use the timings in the Japanese version. It was probably incorporated as a compiled script, hence why we can't change it easily. Alternatively, someone could manually go through the games English audio and modify the Japanese audio to match their lengths. It would be a rather hackish solution though, and would probably sound pretty bad too. The only real solution would be to set up a debugger and patch the kernel to output all files it accesses in the ROMFS, so we could see if it actually accesses an external script file. If it does, it would have to be decrypted and patched either with the Japanese timings, or modified from the Japanese version with the English dialog. In the second case, the issues would probably come from the script layout of the English dialog. Either way, it's not easy if they did it this way, though it's still better than the shit I was going to have to do to get HM-PM2 working. At least there's, ideally, no modifications to the game code here. With HM-PM2, all the main text was stored in the code, and the rest was all pre-rendered 2D objects, that would have to be modified manually. Also, the kerning for the Japanese characters made it impossible to directly translate the lyric files, as the rendering was off, and would usually cause a buffer-overflow, leading to a system crash. That would have taken extensive modifications to the code to fix. That's why there hasn't been any updates on my HM-PM2 thread, because ain't nobody got time for that.
Anyway, good luck to the poor soul that tries to undub this game. All I know is it won't be me.