Of course, when you're recording individual tracks to make a mix, then absolutely use 24 bit, this means that the recorded noise will never be a problem when many tracks are combined. The problem is that, as well as the audio we want to be recorded in the final mix down, there will always be some noise present.
24 bit and a higher than needed sample rate pretty well ensures that noise won't be a final mix problem.
If you really want to, then use oversampling at high rates, although there's not a huge amount to be gained.
The subject is a complicated one, and one, I'm prepared to admit I don't fully understand it, but oversampling can, if implemented well, further reduce the noise floor. Otherwise, very high sample rates just record frequencies that we can't hear, i.e. up to 48kHz.
There's quite a good article about oversampling here: http://www.earlevel.com/main/1996/10/19/oversampling/
We mix at 24 bit at 48 kHz, or 96 kHz (or even 192 kHz for the certifiably insane), then convert to 16 bit 44.1 kHz for playback.
The noise floor is then so low that it should be perceived as silence.
The extra dynamic range and bandwidth are heaven sent in the studio, but not necessary once the mix has been completed.
We can convert the studio, high sample rate, deeper bit depth final mix to CD standard audio safe in the knowledge that it should sound wonderful.
96 dB dynamic range and a frequency response to 22.05 kHz is astoundingly good,
Providing always that the recording engineers know what they're doing. (Not always a given!)
Good microphones, good, low noise, analogue pre-amps and so on.
It also pays to know that the Miles Davis recordings that were referenced were made back in the early seventies.
Even analogue recording techniques back then wre not of the standards that we know today.
It's no wonder they sound "muddy" and lack transparency.
It serves as a reminder to remember that not so long ago when the PSTN (the public switched telephone network) was analogue, call quality was often appalling.
A change to a digital central network called the IDN (integrated digital network) changed the frequency response from about 3 kHz tops, to just 4 kHz.
This was 8 bit samples at just 8kHz.
All of a sudden it sounded as if the person being talked to was "in the next room".
This was the effect of going from analogue to a very simple digital signal.
It's also worth remembering that the top C note of a piano has its fundamental at just above 4 kHz.
Not many instruments have a range any higher than this.
Of course there's loads of overtones and the hiss from cymbals.