It's been a long time since I actually imported a reference track into a project for comparison. The problem, as noted above, is that you're using a mastered track as your reference. That's always going to limit the amount of useful information you can glean from a comparison. Even EQ and vocal levels, which would seem to be the most reliable clues, are affected by mastering compression.
References are still quite valuable to me, though. But instead of directly comparing them to my mixes, I use them to train my ears to recognize what a great mix sounds like on my system. My computer is my primary listening station. Whenever I'm perusing forums, reading news aggregators or going through emails, music is playing.
Right now it's
Sailing to Philadelphia by Mark Knopfler. It's nothing like my current project, an orchestral piece. But it's a prime example of a clean, well-balanced mix and the music suits my mood this morning.