@dalims4
From the very same Web page:
"... our encoders remove the audible ‘digital blur’ that builds up in studio production."
"But a lossless file is just a digital container ..."
So, they modify the digital master first, and then losslessly pack the changed data.
The papers explaining how MQA works claim that there is no perceptual loss. However, technically there is data loss, because original signal gets compressed: they call it "folding", yet it doesn't change general nature of the process.
Moreover, it appears that a signal's digital representation can only be perceptually transparently folded into MQA format if it fits into prescribed triangle on energy vs frequency graph. The energy of higher frequency components has to be below descending line defined in MQA specification.
Correspondingly, experiments with publishing music on Tidal in MQA format uncovered two types of losses: additional noise slightly beyond what mere dithering would add, and rather significant artifacts triggered by signals not fitting into the MQA triangle.
So, like any competently designed compressed format, MQA strives to be perceptually transparent on signals falling within its domain of applicability. Which, arguably, are the most music signals. And in this sense, it appears that there shall be no meaningful losses for most signals.
However, MQA can't be technically called a lossless PCM compression format, because a lossless format has to encode and then decode any PCM file with a supported bit depth and sampling rate in a bit-perfect manner. FLAC is an example of such format. Yet its compression factor isn't high - typically around 2x.
Correspondingly yet again, subjective evaluations of MQA-encoded recordings are mixed. Most files do indeed appear to be encoded in a perceptually transparent manner while being significantly smaller than FLAC. Others reportedly do not. If MQA was consistently the home run it claims to be, we wouldn't be having this conversation.