Remunerating Managerial Talent

Robert Frank is quite certain about executive pay:

Why not limit executive pay? The problem is that although every company wants a talented chief executive, there are only so many to go around. Relative salaries guide job choices...
If C.E.O. pay were capped and pay for other jobs was not, the most talented potential managers would be more likely to become lawyers or hedge fund operators...
The market-determined salary of a job generally offers the best -- if imperfect -- measure of its importance.

So many empirical statements! Are there really "only so many" talented chief executives? Does a salary of, say, $25 million a year really attract someone struggling by on $15 million or so? What on earth makes Frank think that the most talented potential managers would also make excellent hedge fund operators? What, exactly, is the "market" which determines job salaries, and where can I short it?

The whole column is steeped through with the idea that there's a rare and innate quality of talent which can only be found in certain special individuals and which, if married to a company of sufficient size, can result in untold millions of dollars in extra profit. For all I know, this is actually true. But I would love to see some kind of empirical demonstration of it, rather than bald assertions. Given former Top Managers' ability to spectacularly revert to the mean, I do suspect that there might be rather more luck and happenstance involved than Frank would like to admit, and that any old middle manager will, on average, be just as good in the CEO's job as the average spectacularly-remunerated superstar.