as a general approach to my methodology of recording, I usually record at 24 bit, (96 or 192, depending on what rig I am working with) just because I would rather have the programs i am working with have more data points to work with for accuracy than not have enough to dither down to later on in the process.
true, i can't hear the difference between 16 and 24 for most things but just because I can't tell the difference doesn't mean it doesn't contrbute to something in the eventual mixdown process. i agree that the general person doesn't buy for fidelity, still, it doesn't mean I shouldn't put everything of what I can into the mix. As the mixer or engineer, is my job to maintain a consistency even if the other person can't appreciate it that much because they only hear the resulting 44.1 mixdown.
in science and medicine, there is a saying in the lab when working with DNA - just because you don't see enzymes that degrade your work, that doesn't mean they are not there, actively doing the detruction you don't want them to. as such, making every effort to ensure that something comes out in the end as best as it can is a good way of ensuring quality as best as you can provide it, whether it is for music or anything else you work with..
just my 2 cents.