An accurate system model is one that performs as expected and produces the desired results. To achieve this level of accuracy you must have a clear modeling goal. For example, will the model be used strictly for simulation, or for generating embedded C code?
This article describes five modeling techniques that can help ensure accuracy:

* Select the right modeling approach
* Know your system, understand its dynamics
* Monitor for unacceptable results
* Verify simulation results
* Understand modeling assumptions

The full article can be found here:
https://www.mathworks.com/company/newsletters/articles/building-accurate-realistic-simulink-models.html