Is this possible? I don't mean the unchecked OCR version that comes with them for searching purposes, I mean change the actual images used to display the characters, like change 'a' into a 'b'.
So it would be something like extract the files, edit the html that says things like this:
<path id="gl3353" d="M 9 6 C 8 6 8 9 8 18 C 8 21 8 25 8
work out what character each number was, fix the mistakes by changing the number, then put it all back into a mobi file when you've finished.