ranma - well, I mean if you have a stereo track and encode it both before and after downmixing to mono. Most codecs encode stereo as M+S (mid and side) rather than L+R, so in a stereo track with identical L+R channels the side channel is zero and should compress to nothing. Most encoders also use variable bit rate (VBR) so choosing '96 kbps' does not mean it will definitely generate that much data per second, it's just an approximation, and it will be less if the audio is easy to encode. (For example silence won't generate 96kbps of data in a VBR encoder, that would be wasteful). A stereo sound would have to add a lot of data to the S channel, making the file up to twice as large (but usually not that much actually).
The Ogg Vorbis files tend to be very small and good quality, whereas the AAC files are a little bit bigger and sound a bit worse. It's not a major difference, only IE+Safari users get those files, and I doubt many casual gamers would actually notice, so I don't think there's any need to change the way we do it. Again, if you want tight control over how your audio files are compressed, mix and encode them before importing to Construct 2 - if you import .ogg and .m4a files, C2 just copies them in to the project without touching them. It will only encode for you if you import .wav or .flac.