von Mises stress deviation in shells

For verification purposes in designing a large concrete dome I've compared the von Mises stress between the Karamba Utilization component (multiply the output by the yield stress) and the ShellVecResults component (principle plane stress equation). I believe in both cases the results are calculated from the principle stresses at the element centroid. Is this correct for the Utilization component?

For a simply supported beam shell the results match, but for a parabolic dome the values are significantly different. Using a concrete material did not give clear results so an isotropic material was used (a36 steel) with the linear solver. See attached images for results comparison.

If both results are derived from the principle stresses at the centroid of an element why does the von Mises stress differ between the components in the dome geometry case?

Thank you!

- I've checked karamba's solution for the dome principle stresses against the analytical solution and this matches. 

- From the many other discussions regarding shell results I understand that the ShellView visualization and ShellVecResults component can deviate due to interpolation of results at the centroid to the element nodes. I'd be interested to know more about how that weighting occurs but this is not the case described above.