The detection algorithm gets 'confused' because of Esc characters.

The internal viewer does detect the file as text. Maybe the File Comparator can use the Viewer's algorithm? Or at least consider Esc characters as text.
Esc characters occur in ANSI escape codes, in UNIX-like environments as well as DOS and Windows 9x.
All this isn't a very big deal of course, as redoing the compare overriding my default Automatic with Text solves it.