I don't think this is one of those cases. The question we want to answer is whether texts are good enough that humans will not pick up on it. Making the task as hard as possible for humans is not indicative of real world performance once people get presented these texts more regularly.