I just came from a class today where Stephen Roach (ex chief economist at Morgan Stanley) criticized this same gap between reported and actual unemployment. He actually focused on a different worrying number: participation rate. The number of people who qualified as "part of the labor force" declined 3 percentage points, in the same period of time that "unemployment" increased by the same amount.

His point was that the way we model economics data will inherently result in missed predictions. We need to remember this fundamental flaw of modeling when we interpret the results of a model, like unemployment rate.