It depends on how the model was constructed and how it is used. Ideally you expect the vast majority of possible observations not to fit your model. So if they

It depends on how the model was constructed and how it is used. Ideally you expect the vast majority of possible observations not to fit your model. So if they do then that's a strong indicator that you have what you expect. Whereas if they don't you can't be certain if the model is maybe just not quite right.