Economists who argue for stimulus spending on Keynesian grounds understand that GDP is not a perfect measure and that the value produced by stimulus projects may be less than the value produced by ordinary spending. See, for instance, this Brad DeLong post, where he estimates the net benefit of the stimulus and counts the useful stuff produced using stimulus money as being only 80% as valuable as the dollar amount would suggest. Or, as he writes:

The extra people put to work produce $110,000 of useful stuff--that's a benefit. ... However, because we are pulling forward spending from the future into the present--spending the $92,000 now rather than in the future--we are buying stuff too soon, and because the government is all thumbs we are to some degree buying less valuable stuff than we woul ordinarily by buying. Figure a 20% discount--that's an $18,000 cost.