You would expect your models to distinguish the root cause. In your example their are either many infected that do not show symptoms or many related people that don't. In any case one of the models should be off.