It's not that he replaced it, it's that he built on it. The Reverse Three Clue Rule used in his node-based design articles ("if the players have at least three clues, they'll draw at least one conclusion") is a corollary, not a refutation of his previous advice.

The main way it's changed since he wrote this article (and since he wrote his Node-Based Design series, for that matter) is that he distinguishes between clues and leads, which he didn't at the time.