Hi kiwidude,
Again with all due respect, I was a programmer of 40 years so I do know GIGO etc. I know there aren't any magic wands in programming.
Calibre is terrific but with what I suggest and a few minutes programming can be further improved.
When a string contains a " by " where a title and author are expected it would be a small modification to check the right hand side (of the " by ") against the table of author's names Calibre already has in memory and make a better guess.
With this minor mod you will improve the parsing of the available data. Some of the files, in my case, were "txt" so there isn't any meta data to worry about, just the strings in the front of the file. Even if there IS meta data to consider, don't take it at face value, further analyse it. Whatever logic that is already in place can be added to.... :?
In my case this would correctly interprete hundreds more books. This would mean many less books to manually "fix". I would imagine the same would apply to lots of others adding books to existing libraries