How can Mankiw miss the point so much?

Gregory Mankiw has today a column in the NYT arguing that the top 1% have high income because they generate huge amount of wealth for everyone in society:

So, by delivering extraordinary performances in hit films, top stars may do more than entertain millions of moviegoers and make themselves rich in the process. They may also contribute many millions in federal taxes, and other millions in state taxes. And those millions help fund schools, police departments and national defense for the rest of us.

Unlike the superheroes of “The Avengers,” the richest 1 percent aren’t motivated by an altruistic desire to advance the public good. But, in most cases, that is precisely their effect.

The use of the actors and basketball stars to make his point is clever but the column misses totally the point and ignores convincing research that points into a different direction.  Most people recognizes that some levels of income disparities are justified: unique talent should be rewarded.  The question is how handsomely should we reward it?  Is there a reason Robert Downey should get paid 55 million for his role in "The Avengers"?  Would he be unwilling to act and get famous if he was receiving just 10 million?  5 million?  Robert Frank from Cornell and others have argued convincingly (see, for example, this book) that most people care about their comparative income positions and that inequality forces everyone to aspire for more consumption and spend more on luxury goods.  The world would be a much better place if talent was rewarded... but at lower income levels than now.

Mankiw is in particularly shaky grounds when defending astronomic rewards in the financial sector. How he can defend with a straight face after the global crisis that the role of hedge funds and banking managers is only to allocate credit to the most profitable and important activities in society is hard to believe.  Rents and not profits from competitive markets  is what the financial sector is too many times allocating.