If I felt that one could reliably attribute everything that went wrong or right to the individual in question, I might agree with you completely. It's not clear to me that even highly-paid superstars can necessarily produce results that justify their salaries, simply because there are all too often a lot of exogenous factors that one can't control, in addition to facilitative factors you might have had nothing to do with. Government is superimposed upon the relentless trudge of history, world markets, natural and man-made cataclysms, people's personal lives, etc.. I don't know that salary is going to override an executive's kid or spouse coming down with leukemia, or supercede a massive change in the labor market or economy. Stuff happens, and money seldom makes it go away.
I'm not throwing my hands up in the air, but my sense is that truly competitive salaries work when a) one can directly and consistently link performance to individuals,.and b) you have enough flex in the budget that you can afford a little inaccuracy in that first thing.
And, lest I digress too far into a redundant pay-for-performance debate, many writers concur that pay-for-performance requires certain circumstance to work well, and those circumstances are only rarely achieved within the public sector context. One of those circumstances is that there has to be some degree of general parity in the awarding of salaries, simply because people compare their compensation against other people, not against their own individual performance. So, what a high-performer earns in agency X better be what an equally high performer in agency Y gets for the same stellar performance, or else you're going to keep losing people in the agency where they feel undervalued. And once you get into contests about which agency warrants a bigger budget to attract and hold high-performers, you tend to invite problems.
Call me a hopeful skeptic. I'd certainly like it to work as you describe, but I have my doubts that it will.