Frost, well here is the "sort of" . . . FLAC uses linear prediction to convert the audio samples to a series of small, uncorrelated numbers (known as the residual), which are stored efficiently using Golomb-Rice coding. This is how it reduces the file size over 50%.
First of all nothing makes 100% perfect predictions ever, if something did I'd be using it for buying stock and playing the lottery LOL. Tangent comment: there are large computer servers now going in service located very close physically to the stock exchange's computers so that they can do millions of buy/sell actions per day using the advantages of linear preditions and low latency (low ping time) to make money. Given that and the extremely high cost to lease them I would assume that linear prediction has a pretty good track record statisically.
Secondly there is a bit element which you refer to and a time element which some use the word "jitter" when talking about this discussion. Here are a couple of web debates touching on the issue that may be of interest: The Well-Tempered Computer FLAC versus WAV
Even with a null test, one would have to ask how wide the time element is set in that test and how sensitive to this time element (jitter) the listener is given a good monitering chain that would reveal differences in the first place. Just like when doing blind testing of clocking options for our setups, using a stereo file with lots of 3D info in it would be the better test reference song given if the timing element is off it tends to move 3D sources to the up front center position (smearing the timing cues our ears need to locate a source's exact position in the sphere around our bodies). The simple answer is get your hands dirty with your gear and see if you hear a difference and if so does the difference matter.