More broadly, getting the metadata right is the core of the utility here, and editing it manually is the last-ditch way to do that. You simply must make that process as fast and efficient as you can, and copy/paste from the document itself is much faster than retyping, obviously.

Seems to me that metadata import/management is *the* (potentially) distinguishing feature on which Mendeley eventually succeeds or fails. While users will give you a break for awhile, in steady state the "acceptable" bar here will be very high, and doing it badly will eventually be fatal.
Closely related is the ease with which users can fix incorrect data. I just imported my library of several thousand documents -- hundreds of them should have been read correctly but were mangled, and hundreds more are talks or whatever that I'd like to catalog but don't expect to be automatic. I'd grit my teeth and fix them if it were slick/fast, but it's slow and clumsy. So, give up, or wait and hope that the UI gets better?...