Don't entirely rely on Content-Type headers to identify PDF attachments
Some systems tend to put nonsense into Content-Type.