Do you know whether anything like this been formalized and proven?
But let me give an example that would make it more concrete. Suppose you want to build a computer that simulates the behavior of n bodies in gravitational interaction in space. It's obvious that with a digital computer, there would eventually be discrepancies between the actual system and the simulation (which would only grow). But now consider this: suppose that instead you want to do the simulation by building a "computer" which is nothing more than a copy of the original system (i.e., the "simulation" is composed of another set of n bodies in space). Is it possible to have this analog "simulation" more accurate than a digital simulation? The answer is no, because even supposing a true continuum of positions and velocities exists in the first place, in order to reconstruct the system precisely, you would need to pass an infinite amount of information to the "simulation", which is impossible.