While these ports sound good, unfortunately, they have several issues on the MML side that prevent them from being accepted.
The most noticeable issue is the lack of optimization. While there are a few loops here and there, there are still a lot more that could be added, along with increasing the amount of label loops so the same loop isn't redefined, and extending certain loops so they don't end without including all the note data that repeats, and so on.
Other than that, there are also a few minor issues that should be looked at as well.
-Skyscraper, Memories, and Premature Death all insert with #default, so it'd be preferable to use #default instead of #optimized for them.
-The global volume for Memories and Premature Death could be lowered a little bit, since while the songs themselves are quieter, the samples are pretty loud. Similarly for Skyscraper and Song and Dance, though less so.
-In Strong Fortress, $F4 $02 is placed at the beginning of channel #0, but since there is no loop point (/) this will cause it to toggle on and off every time the song goes back to the beginning.
Fixing up these issues should make these ports good to go.
Also, one last note, which is more a writing nitpick rather than a rejection reason, but the repeated use of h commands can make the .txt a bit harder to mentally parse when reading over it, and might even make you not notice some potential loops. As well, stuff like h-18 could just be <h-6, since any increment of 12 semitones corresponds to a full octave anyway.
Here's a visual recap of some of these issues. It highlights some of the issues across various ports, but just because one example only comes from one port, that doesn't mean it's not present elsewhere. It's nowhere near comprehensive, but it can hopefully kickstart what you should be looking for.