Suppose you flip a coin on 100 candidates, one of whom is qualified. You'll get about 50 "hire" answers, and the odds are 50% that one of those hires is actually qualified.

So no, a coin flip is far worse.

The article is referring only to employees that were hired and were also given the GMA, so qualification was not an issue.

However, using a coin flip rather than a GMA score on a prospective employee would imply a score of either 0 (missed 'em all) or 100 (perfect), and thus would likely have 0 correlation with the actual employee performance, rather than .51.

So your conclusion is right even if your analysis is not.