Come on Economist, realize that we're plotting averages of averages and then fitting that average. Not too scientific. Let's break down the underlying formulae, solve to carry through the dispersion of the underlying data, and then add the confidence bounds on the sides of that bestfit. If they did, we'd see much lower confidence around those 2300-2500 hour points.
But wait, we've got an R2 of 0.2! Let's keep in mind that at best, high hours of work are not the dominating factor that killed you.
