The biggest problem with the theory that the frequencies you cannot hear affect the ones you can (anything above 20KHz). is that hardly any microphones or monitor speakers reproduce these frequencies anyway, so they are never there to perceive - whatever sampling frequency you use.
The only “sound difference” in high sampling rates is usually that of the D/A converter (quality of the aliasing filters), and that is in the end users CD player or computer (read - not very good).
How ever it is easier to design a cheap (poor) D/A running at 96KHz than one at 44.1KHz, as at 96KHz, you can heave really rubbish aliasing filters that do their job without affecting the audio.
So maybe the sampling rate is best higher for all the wrong reasons.