OK - from my experience the main reason the mix engineer wants uncompressed stems is because :
a) he's usually a post production engineer and doesn't feel comfortable mixing the whole track, or doesn't have time in the session.
b) he can't uncompress things, so this way he can add the "suitable" amount of compression at his end once the voice and effects are in.
c) The voice is the most important thing in TV mixing , no room for debate, the agency will tell you so. The voice pretty much fills up the 1 - 4 K area, so simply pulling back the stem with the most energy in this are can clear things up and make room for the voice without having to pull all the mids out of a full mix.
As far as your approach, the post production mix engineer is unlikely to compress all your stems individually, he will most likely buss them all to a stereo group and compress that. You could work this way by routing to your subgroups then bussing them out the main stereo out and using a light compressor over the whole mix for monitoring the effect it has.
As you become more confident/proficient at mixing the other way you could do it is to get a file of the voice over and run it in your project as you mix. This way you'll hear where it sits and be able to eq around it. Then mute it and mix a stereo master of your track, and don't supply stems.
As I say, it's up to you to be confident enough to take this approach. I don't supply stems because no post engineer is going to have my perspective on how the music should be mixed. It also leaves you open to having your music edited without prejudice by some moron who's not a musician, be it post guy, director, agency producer. "Hey that sounds really cool with the bass taken out and the drums moved out of time...let's go with that one". I had a full symphony score for a film that had a few cues absolutely butchered by the director in the final sound mix because I was out of town on a job and couldn't be there. I didn't know about it till the preview in a big cinema and I wanted to die. To them it was just background.
Having said all that, if you have to mix stems use the compressor across the stereo sum of the stems and mix it so all the faders of the stems are flat, all set at unity. At least a f&ckin monkey could get it right then. Well most of the time.