I tried to mirror a website which uses application/xhtml+xml as Content-Type.
No other files but the robot.txt mentioned in the HTTP-Header was transfered.
I tried that locally by creating some small sites and marked them using
.htaccess option "AddType application/xhtml+xml .xhtml" with the correct
Content-Type. It showed the same problem, but changing it back to text/html
fixed it.

The urls didn't include .xhtml file ending, but adding it to the url didn't
changed anything.

Downstream bug report is <http://bugs.debian.org/563691>