I think you're missing my point too. The models are only inaccurate temporarily. They're still quite valid for predicting longer-term trends.
The reason for this is because of the cyclic nature of the ocean warming. ENSO, AMO, PDO etc are all natural cycles that shift heat between the ocean and surface. While heat is being transferred to the ocean (as currently), the models will over-predict surface temperatures. When the cycle reverses, heat is transferred from the ocean back to the surface - and the models will under-predict surface temperatures.
Because heat isn't created or destroyed, only moved around, the net effect of these cycles is zero, and does not affect the longer-term warming trend that the models show. If you take a longer-term running average of surface temperatures, they still come very close to the models' predictions.
Climate scientists already know that the models won't match the short-term noise created by these currently-unpredictable cycles - but that's OK, because they're not intended to. It's only uninformed laymen that insist that any short-term mismatch is a "failure" of the models, rather than simply using them for longer-term predictions only, as designed.