Quaternion models have been developed for both direction of arrival estimation and beamforming based on crossed-dipole arrays in the past. However, for almost all the models, especially for adaptive beamforming, the desired signal is still complex-valued and one example is the quaternion-Capon beamformer. However, since the complex-valued desired signal only has two components, while there are four components in a quaternion, only two components of the quaternion-valued beamformer output are used and the remaining two are simply removed. This leads to significant redundancy in its implementation. In this work, we consider a quaternion-valued desired signal and develop a full quaternion-valued Capon beamformer, which has a better performance and a much lower complexity and is shown to be more robust against array pointing errors.