I am careful to point out in the article that the further into the future we are trying to forecast particular deliverables, the less empirical (and agile) we are being. But the core purpose of this article is to offer an alternative to story points which, like any estimation methodology, is flawed if the team is not doing the things you describe.