Failure in metals is strongly correlated to the macroscopic stress state, in particular to the stress triaxiality. Fracture specimens are designed to probe a range of stress triaxialities so that a failure criterion can be fit to the observed experimental behavior. For metals exhibiting weak plastic anisotropy, analytical functions can be used to relate the external load and displacement to the stress and strain state in the failure zone. However, for materials exhibiting strong plastic anisotropy, such relations are inaccurate. Strongly anisotropic materials have shown seemingly anomalous dependence of failure strain on stress triaxiality. In this work we re-examine existing failure experiments performed on magnesium alloy AZ31B that were designed to interrogate the dependence of failure strain on stress triaxiality. Finite element crystal plasticity simulations are used to more accurately relate external load and displacement to the stress state within the failure region, as well as slip and twin activity. We then examine whether or not existing failure relations can be used or modified to fit measured behavior.