The 64X oversampling is cutting the existing data into smaller sizes. Its the equivalent of modulating an audio signal with a high frequency carrier like in AM or FM. The decoded OS signal is still the same in shape as the original. An upsampled signal is a close approximation of the original, but not the same. The extra 0s added to stretch a 44.1KHz to 192KHz make a big difference. Get a digital storage scope and do some tests with waveforms through both methods of sampling. The theory is getting in the way of the practical outcome.