There aren’t that many audiobooks in the MB database, so I’m guessing you are grouping things in Picard and then hitting ‘Save’? Rather than using ‘scan’ or ‘lookup’?
If that’s the case, it is applying the default script to the existing tags on the files. For instance, if the narrator is in the ‘Album’ tag field, or ‘Artist’ field, it will probably use it to create folders. Is this correct?
Picard isn’t ideal for that kind of thing - it works best when you can look up the release in the database, and save the correct tags from that, and then it applies the script, making sure that everything is uniform.
That said, I’m sure people will take pity regardless 
But you have to give a better indication of what steps you’re doing and what you want your files to look like.
Or you might find a perfect script that already does what you want here: