Using for instance tsmuxer's "append" to join multiple h264 streams into one mts-container is absolutely fine, as long as the number of encoded frames per part is a multiple of the GOP-size. Then it'll be a "clean" h264 stream. If it's not, certain hardware decoders MIGHT have a problem, although I doubt that anybody will actually run into one that can't handle it.

I had the same problem with a couple of video projects. Some frames just didn't make it error free into the final master lossless file. Problems like these are, in my opinion, the reason why it's never a bad idea to render an image sequences out of After Effects in the first place, and only go to a lossless codec "master archive file" after an initial test encode. Although, I never bother to, I just keep the image sequence as the master. It's easier to archive to DVDs too, I just divide them into chucks of 4.2GB and use the rest of the disc space for parity (PAR2), just in case the DVD fades away after a few years.