If you have a system of ODE that gives rise to a chaotic system, you can easily find that the solutions either explicit or implicit are drastically different from real-world results. Yet, this does not mean the model is inaccurate, only that you didn't pick the correct infinitesimally precise initial conditions to perfectly match physical reality.
In practice, how can you distinguish between an accurate model that is chaotic and a model that actually is inaccurate?