People wildly underestimate how long certain kinds of tasks will take. This is usually regarded as a Bad Thing.
Imagine how this plays out if you're a decision maker, deciding whether to do highly ambitious project X. If you knew ahead of time how long X would take, you would never do X. If you don't start on X, there's a zero percent chance you ever finish it. But because you have an unrealistically low estimate of how long X will take, you decide to do it. So now there's a nonzero percent chance you finish X. Once you've started X, the project has a nonzero chance to stay alive, even though it ends up using massively more time/cost than planned. As a decision maker, you don't shut down project X when it overruns its budget for a couple reasons: Partly due to sunk-cost fallacy, and partly due to continued underestimation of how long the remaining parts of X will take.
The author's point is that this underestimation is actually a Good Thing: If we could accurately estimate time costs up-front, fewer highly ambitious projects would be attempted (and therefore, fewer would be completed).
Imagine how this plays out if you're a decision maker, deciding whether to do highly ambitious project X. If you knew ahead of time how long X would take, you would never do X. If you don't start on X, there's a zero percent chance you ever finish it. But because you have an unrealistically low estimate of how long X will take, you decide to do it. So now there's a nonzero percent chance you finish X. Once you've started X, the project has a nonzero chance to stay alive, even though it ends up using massively more time/cost than planned. As a decision maker, you don't shut down project X when it overruns its budget for a couple reasons: Partly due to sunk-cost fallacy, and partly due to continued underestimation of how long the remaining parts of X will take.
The author's point is that this underestimation is actually a Good Thing: If we could accurately estimate time costs up-front, fewer highly ambitious projects would be attempted (and therefore, fewer would be completed).