I'm a little skeptical of the attack on the Nyquist theorem, given that it is a completely mathematical proof that, when sampling, one can absolutely reproduce the original signal if the digital one was sampled above the Nyquist rate. No one has yet debunked Nyquist, not even analog electronics design engineers and professional musicians. I would be open minded to attempts to the contrary, so if you think you have debunked Nyquist, then please show me where the fault lies.
Now, there is definitely a good point to be made about the limiting nature of quantizing all of the samples artificially to 16 bits. Obviously this is the one case where analog recording shines over digital recording, in that analog recordings have headroom limited only by the physical properties of the recording medium (magnetic susceptibility of tape oxides, for instance).
There might also be something to the argument that 22 kHz is too low for the maximum frequency in audio signals (which determines the Nyquist frequency of 44 kHz). I have heard some say that frequency harmonics above 20 kHz DO in fact play a part in shaping noises, even though most human ears cannot directly discriminate between a 20 kHz tone and, say, a 22 kHz tone.