No mention was made as to whether or not any performance was above or below expectations. You are assuming it is below expectations in every case. The Carlsen event is a possibility since he didn't win it by much. But, the Fischer win over Spassky was a dominant win most likely above statistical expectations. Yet, Fischer lost rating.

Thus, the problem is a problem within the rating system. Every instance I have found where a player lost rating when winning an event with the Elo rating system showed my rating system giving them a boost in rating.

Is my rating system correct every time it disagrees with Elo? It might not be, but it is more powerful than Elo, so it is likely right over Elo.