The robust Capon beamformer (RCB) is more robust than the standard Capon beamformer when there is uncertainty, as in the case of a sensor location error or an array model mismatch. However, the RCB suffers from mutual coupling between the array sensors, which is difficult to neglect in practice. In this work, an autocalibration algorithm is described that improves the RCB by recursively calibrating the uncertainty due to mutual coupling. The proposed method is better able to estimate the angle-of-arrival and power of the signal of interest than the RCB, as shown by numerical simulation and practical analysis of experimental data obtained in an anechoic chamber