The problem with refactoring is the cost. Its rewriting code to no client perceivable benefit. The retesting, the inevitable new bug introduced during deployment.

As developers I think we need to assume the cost of refactoring and factor that into time estimates for our work. What the author is in effect saying is that we should factor 20% extra into our estimates for rewriting code. Which is fair, I just don't see a need to specify a day for it. Just put it in your project plan.