Well, things are going better than I thought they would. I've almost managed to reliably parse to the end of a save file - by reliably parse I mean I either understand what the data means, or I can figure out how much of it there is and skip over it. I've tried not to skip over too much stuff and I only skip 30% of the file at the minute. Because there are a lot of repeated sections I can quickly reduce this amount by figuring out small parts, which is what I'll be doing soon.

I've also figured out a few things that were confusing me before, such as visibility information. I'm also writing a specification document for the save format that I'm updating as I go along - it probably won't be 100% accurate since I'm only working from observation, but it should be useful anyway.