I'd say that there's a reality (whatever it is; some phenomena that can be observed in whatever ways), and there are models of it.

Unless the reality has very particular properties (some kind of regularity / redundancy), every model of it is a simplification, so it has discrepancies. Even if the reality is regular and compressible perfectly into a model, it's rather hard to come up with such a model, and to prove it's the right one.