>PCM is also a lossy compression due to the quantization step
Wrong, especially today. Modern ADCs use oversampling to push quantization noise into the inaudible range, and then filter it out before decimation to standard PCM.
Because the end result is standard PCM, the quantization can be only worse, not better.
Oversampling ADCs push a much greater quantization noise into the inaudible range, and then, by low-pass filtering, reduce the quantization noise to the level of standard PCM.
Oversampling ADCs are not better, they are much cheaper, because 16-bit or 24-bit quantizers with enough speed and accuracy are extremely expensive.
Oversampling, i.e. sigma-delta modulation, in both ADCs and DACs, allows the use of much cheaper quantizers with low resolution, of only a few bits, or even of only 1 bit, and of much cheaper filters, which do not have to be very steep, without degrading too much the quality in comparison with real PCM conversion done at the Nyquist frequency.
Wrong, especially today. Modern ADCs use oversampling to push quantization noise into the inaudible range, and then filter it out before decimation to standard PCM.