Understanding the sensitivity of energy-maximizing control for wave energy converters (WECs), to various model errors, is crucial for application. Many advanced WEC controllers, especially model predictive control (MPC)-like controllers, require estimation and prediction of wave excitation force (WEF). However, previous studies only focus on the controller in isolation, without considering the error coupling effects when a complete estimation–prediction loop is involved. In this study, it is revealed through numerical analysis that the complete MPC system has sensitivity behavior completely different from the isolated MPC; under certain model errors, the system can become particularly unpredictable, exhibiting potential instability and self-locking phenomena, which cannot be observed from the examination of control sensitivity alone. Meanwhile, different tuning options for the WEF estimator and predictor are examined, where the accuracy–robustness tradeoff is shown to be critical for performance amelioration under errors. Based on the analysis, this study challenges the widely assumed “separation principle” of WEF estimation/prediction and WEC control, highlights the importance of incorporating a complete estimation–prediction loop in sensitivity examination, and draws practical guidelines for WEC control application.