The point about mixing on anything - not just digital is to maximise the s/n, so the gain staging thing is important - but if you are throwing signal away, and then have to boost the resulting file, then you got it wrong, because even a big dynamic range could have been bigger, if managed a bit better. You need to experiment with your system to see where it's limitations are. I mentioned that the actual level of each track doesn't have to be exactly the same, because in a multitrack CD, there is also a balance between the loudness of each track. If you have, in Cubase for example, just a few hot tracks, it's easy to have these summed in the mix bus and need the output level reducing to prevent the red light coming on. With lots of hot tracks it happens very easily. You can work with the channel faders lower, but with analogue and digital mixers this is a pain because you squash the useful operating range of fader travel into a lot smaller area - making delicate adjustment more difficult. If your recordings are too quiet, compared with others you have to inspect, then shove the master up a bit and do the mix again (presuming you recorded the fader movements?)
In the recording and mixing stage you are maximising signal to noise, and finding a working style that does it for you. My method is to have the nominal position of the faders at around the usual ¾ ish position for the key tracks, and push or pull for balance from that position. If you discover you are pulling back (or pushing) more and more tracks, then I reduce or increase them all, and do the balance again until I have a convenient working layout. If the track really makes use of the dynamic range available, then I will bring up the master and watch for headroom - which the cubase meter shows you quite easily. The resultant tracks are quite loud, and the relative balance between songs is restored later on. Once I have a .wav file, it's not good to have to process that file to make it louder. Quieter is easy!