The blind tests aren't even the end-all here. The science of human perception limits and characterizing the error are the two fronts to look at. We've long passed recreating what a human ear can hear. We have targets for frequency range, dynamic range, and group delay. Hit those targets and you've won. There is more on the art side of things to advance, namely in simulating directionality, but that isn't what the Tweeter salesman is usually selling you.
The imperfections of the audio chips in old consoles are easily audible to humans. If the goal is accurate recreation then assuming the audio chips and driving circuitry is perfect is insufficient.
Video recreation is still far below what the Human Visual System can detect. We are not masters of the universe yet.
