Comment 5 for bug 542755

Ok, I have to admit you're right. I was under the impression that pdfshuffler's behaviour was the standard one, i.e. all or at least most file dialogues filter only on the extension and only offer the "all files" option to get other files. But I was wrong, the behaviour you expect is indeed the standard one. Also my comment on speed was not really fitting: In my case (folder full of PDFs with correct extension) it won't make any speed difference, it would only rely on the slow detection if the file has no extension.
The only thing that would remain is point 1) from your description but this would really be standard behaviour, e.g. eog would list a file with .jpeg extension regardless whether it really is an image or not.

To conclude, this issue is an upstream bug, I'll report it in the upstream bug tracker. The patch is actually trivial, in the /usr/bin/pdfshuffler file, replace the line
         filter_pdf.add_pattern('*.pdf')
with
         filter_pdf.add_mime_type('application/pdf')
(Note that whitespace is important for python, i.e. you have to have the same indentation as before).

Sorry for my negative earlier response, I learned something new today ;)