Oh dear.. Sighs
The whole point of stems is that the dial, music and fx stem should sum to make the dm&e, to which you dip under the VO stem to make the final mix..
I reversion hundreds of hours of tv a year for my friends at AETN, and if I had a pound for every time that this wasn't the case, I would be heliskiing in NZ now.
The music stem should be therefore be the correct level under the sync (I.e dipped) for interviews and soundbites, but the dipping under the VO is done by the dme fader,, this is because America - yes you! there are other languages in the world that are different lengths (and I count British amongst the other languages) and to fight a dipped dme is a pain.
For most tv shows, mix the m&e fist, the international version is the one that sells and often makes the production the most money, and then go back and final mix for your territory.
This really isn't hard and and it is surprising how many people get it wrong and fall foul of the golden rule of post "don't pass [bleep][bleep][bleep][bleep] on"
Most international versioning sessions do not have the time or the money to rebalance your dme, so you make no friends and are not helping yourselves if you provide a dipped dme and you and your post house will rapidly get a shirty call from the production company.
Incidentally, although I am often found being ironic and perhaps irreverent, when I made a quip about foreign languages above, I was actually being quite serious, most American tv shows are re-voiced in the UK as our style of language and also our style of mixing is quite different. I digress.
here's a hopefully helpful video from Pro-Tools-Expert where Mike Thornton and I discuss how to make undippeded M&E mixes, and how to lay out your templates and why.
It is not intended to teach seasoned pros how to suck eggs (before you even start), but if you are not sure or have little experience at this, it will guide you along your way and prevent you from making silly errors.