I'm sorry for not answering immediately but I needed some time to study this case. In my previous answer I wrote:
In general this is based on superposition of axial load and bending in one direction plus axial load and bending in the other direction minus influence of the fact that axial force has been considered twice.
The investigation suggests that the reduction of 'doubled' axial force is not done entirely corrects and its influence for the situation as yours is excessive. For the tension in the anchor that is close to its resistance for unidirectional bending this results in going over the tensile strength when additional bending in the other direction is added. This in turn causes the rapid increase of the verification ratio.
Thank you for your bringing this issue forward.
This issue has been addressed in SP3 for Robot 2013