That’s interesting. If accurate, it might be because manual based methods are designed to work as a whole, while fragments of certain methods only work in the context of the rest of the manual. I’d want more independent confirmation to trust the result myself though.