Think about it this way. You take a crisp photo, blur it a bit, then sharpen it, then blur it a bit again. The first and the second image will not looks identical. The first blurred image is the simulation, then we “sharpen” it by observing it at a subatomic level, then “blur” again by calculating up from the subatomic observations. The calculations and the “reality” would not match, but only because the quantum data was generated complexity not necessarily exactly representing the original.

So, if the base of the reality is designed around something on an atomic level, when you zoom in past a given level, it would stop making sense if the program was designed to continually add complexity as it is studied more and more, perhaps in the effort of discouraging people form thinking that they are in a simulation :)