If they went out of their way to make base65536, I'd be surprised if they didn't evaluate at least a few compression schemes. It's mentioned in the readme that

If they went out of their way to make base65536, I'd be surprised if they didn't evaluate at least a few compression schemes. It's mentioned in the readme that they are using RLE so it seems reasonable to assume that their data tends to have long runs that compress well with RLE.