It suggests that the awesome people with backgrounds in crappy jobs got out of the crappy jobs because they're awesome, rather than being awesome because they did crappy jobs.
If you only look at the set of people who work in white collar jobs, I would wager those who previously had blue-collar jobs exhibit some positive characteristics.
I don't think that the set of all people who have had or currently have a blue-collar job is particularly relevant to the argument, other than as a backdrop for speculation.
Correlation is not causation. You could just as easily interpret his statement as P(CJ|A) > P(CJ|~A): the conditional probability that someone had a crappy job at some point in their life given that they're awesome is higher than the conditional probability they did given that they aren't.
