. So what you hear is 16/44.1khz signals at 20/48khz resolution. Because more bits are used, and a slightly higher sample rate, you get lower noise and greater dynamic range.

This can only be true if the master had the 20 bit depth. If your source is only 16 bits, that is all you can hope to ever get. But I would defer this to better experts in th efield, beyond both of us. Sorry.

But 16/44.1khz contains signals that are heard, and unheard so in theory because of the masking effect of louder signals in relationship to the softer unheard signals, you are not hearing more detail.

Yes,
That is exactely why perceptual coding works so well.