I like this port, it sounds nice, but I have various issues related to optimization:
You should label r1 loops, and also loop them as r2 instead of r1, for example if you have [r1]6 then it's better to do [r2]12, this saves more bytes because r1 is treated as r2^2 internally.
Now, there's also loops like [c]2, [cr]2, [cr]3 and similar ones which are too small and end up increasing insert size instead of decreasing it, so undo those and leave them as "raw" notes. This also means that you will have to redo most of channel #6 but this is still easy to optimize so I doubt you will have issues with that.
Among other stuff, there's also some simple things you missed like this:
And lastly, in channel #1 it would be better if you undid loop (0) and label-looped each line independently:
(Still has loop (0) in the image but you get the point.)