I just tried to use the estimate error tool under Model / Results, but it seems like the error estimate neglects the rules which are used to average stresses between elements, such that the error on all elements which form a T-junction (e.g. between web and flange elements of an I-beam) have the same error estimate at the common node.
That's pretty worthless, because of course there will be a difference in stress between web and flange, for example. In the case where the flange is loaded in lateral tension, and the web has no load, the error (as currently measured by FEMAP) would be 100%, whereas the actual error is 0%!
Please improve this to obey the stress-averaging rules (adjacent angle and also thickness or material changes).
Also, the contour shows a maximum value of about 103 (pic above is limited to 10 (10% error)). I thought that it should be limited to 100, because it is a normalized percentage measure.