Saturation is a very common nonlinearity in control systems and may produce serious performance deterioration or even loss of stability. To cope with saturation, several anti-windup (AW) schemes have been developed over a long time. Unfortunately, they are based on the assumption that there is a static nonlinearity between the output of the controller and the plant input, which, in many situations, is not the case, because of an actuator dynamics. Against this background we provide a design procedure for the design of the AW-compensator that guarantee stability of the observer based anti-windup to face unmodeled actuator dynamics and guarantee a certain level of performance. This mixed performance method is later extended for systems with unmeasurable actuator outputs by the use of an unknown input observer (UIO). The effectiveness of the presented algorithm is demonstrated on an engine test-bench simulator.