The one extraordinary employee you throw away could easily outweigh the benefits of increasing job performance on average.
Drug tests are actually related to job performance as well. If you correlate drug test pass/fail rates with supervisor ratings of job performance within companies that use drug tests but don't hire on them, there actually is a correlation - people who pass drug tests perform better. It's also easily legally defensible, which makes it a no-brainer for most hiring organizations.
Personality actually is correlated with job performance as well, and is job-related. The most empircally-supported personality trait is conscientiousness, which is a person's general tendency toward organization, orderliness, and meeting deadlines. Guess where that kind of tendency is useful. If you already like being organized, and your company wants you to be organized, you're probably going to be better at it than someone who doesn't like it.
Also, if you are conducting a major hiring effort to add 100 employees, your assertion probably is not true. 100 slightly above average employees will still better produce better overall organizational outcomes than 99 randomly hired employees and 1 exceptional one.