I think the misunderstanding is based on the fact that we need >2 samples related to the highest frequency that occurs in a waveform. Natural sound wave has rather transient character and its frequency components go behind the 20 kHz audio band “limit” , so the accurate record and reconstruction needs appropriate sampling frequency. 96 kHz is usually enough, 44.1kHz is definitely not enough to get accurate sound record. Same applies to playback. Please note I am saying nothing about audibility.
Scientific link on audibility
Scientific link on audibility
Last edited: