1。 Is your model an accurate model? If there are some idealized abstractions in it, it is normal that they are not completely consistent.
2。 What is the error of your actual test results? Any system has measurement errors.
If we study from these two aspects, we may have an answer.