So essentially the source code is LLM's best guess what the original might look like, as long as it compiles the same binary the goal is met.
Interestingly; the way that MSC51 and LINK 5 lay out the binary depends on identifier names. There's also length restrictions due to memory limitations. So although the names aren't going to be historically accurate; the possibilities are at least bounded!