This is a great question. The truth is the compensation consultants do some of both. The form of this question seems to presuppose that "accurate" and "worse" are different outcomes. In the world of compensation they often are the same thing.
If a company determines that it must pay their CEO in the 75th percentile in order to get the quality of executive that will best serve the needs of their shareholders then they will look to market data from peer companies (easily accessible for public companies) and determine the 75th percentile. This sets the basic pay level. Easy peasy except for a few little things...
If everyone decides they need an executive in the 75th percentile and everyone is looking at the same data and the data is updated every year, then the 75th percentile will also rise every year. Perfectly accurate and inarguably worse.
If the peer groups selected are just 10% better than a "real" peer group of similarly performing, sized, etc. companies then the compounding factor above is made even worse. Still totally accurate, but not any better.
If the stock market drive upward at a rate faster than initial predicted the compounding factor can be even worse (You Get 2 Shares and You Get 4 Shares and So on and So on…).
But, if you take a look from another perspective you may see a different picture. Investors are all about making money. If I told you you could make $1000 by paying someone $500, or you could make $400 by paying someone $300, you would probably choose to pay $500 and get the bigger payoff. In this case (and this perspective) the higher pay is better. It is also accurate.
All of that being said there is some academic research on this topic. A good start is here. Page on northwestern.edu