If you were optimizing for happiness only and had no ethical concerns about methods, then yes, a very flat or close to flat distribution is likely optimal.

Like you, I believe that there are things more important from the perspective of society than mean happiness.