FileDataSource:
Javadoc for getContentType():
    /**
     * This method returns the MIME type of the data in the for of a
     * string. The semantics of this call are to first search the
     * per instance MIME Types registry and then check the system default
     * registry returning the result of the first successful query.
     *
     * @return the MIME Type
     */

Actually, this lies. The per instance FileTypeMap if any, is queried, and the result
returned. There's no code that falls through to the system default registry when a
instance regitry is provided. In fact,
the api for FileTypeMap.getContentType doesn't encompass the idea of an 'unsucsessful'
query, since it returns 'application/octet-stream' when no mapping is available.

I assume that it's the documentation that should be corrected.

(While your're at it, note: "...in the for of a string.")