Hello, to your questions:
1. you have to define all the functions in the NLCem2User material dialog. Please read the corresponding part of the ATENA Theory Manual and do not forget that the localization strain value is very important - simply said, it controls when the switch from strain-based model to absolute crack opening (or plastic displacement) occurs.
2. I am not sure if I understand your question. If you wish to define an analysis stop criteria, I am afraid currently only convergence errors are supported.
If you ask how to recognize the failure mode, basically do the same as you would with a real structural member or test specimen, i.e., look at the plastic strains in reinforcement near peak load, crack pattern and openings, etc. Just it is much easier in the Postprocessor to look inside the member/specimen...
3. please send us the model along with a description which units or values are incorrect, such that we can reproduce the problem.
4. the "rigid" material should have E modulus about 2-10x that of steel (as you have already understood, a too high elastic modulus can lead to numerical problems). Make the plate thicker and mesh it with large elements if you need to distribute the load more evenly. In some situations, it can make sense to apply the load in 2 points (and sum) instead of a single point.