Because the users who were less clearly fake would still degrade the experience of the rest of the users. To use an analogy, consider currency counterfeiting. The government doesn't just look to see who is spending lots of cash without a job because it's a much harder problem than making the bills extremely difficult for the layman to forge. Same principle here - making the token extremely difficult to forge is the easier route. You don't catch 100% of the bad actors in either scenario, but why not use all of the tools available in your toolbox?