We did some work last year using the Simulate 2.0 elastoplastic capability and evolved our own workflow. Your mesh problems will need to be resolved before you can proceed to any kind of analysis. It makes no difference whether this is an elastic or nonlinear analysis. For models that use a lot of idealisations this can be time consuming as you will know.
Based on our limited experience of Creo's nonlinear analysis capability I think you will need material models that approximate the materials you plan to use in the 'real' hardware. This stumped us originally as PTC dont offer any and there are issues around building material models that make for a stable non-linear analysis. We found that the best place to start is with authoritative stress-strain curves (must be true strain) for representative test pieces of the material you want and finesse the data to get a good curve-fit in the material modeller in Simulate 2.0.
The analysis should first be run as a linear analysis to check convergence and freedom from modelling errors. If that works it is safe to substitute nonlinear material models where desired and re-run the analysis using conservative load and approximate convergence settings. If that works, more refined versions of the analysis can then be run until you have obtained the insights you need. If it wont work, the material models probably need more attention. As always, build up more complex analysis models in modular fashion, dont expect to do everything at once.
We also recommend using User-Defined Output Steps for complex non-linear analyses as it can greatly help analysis convergence and (in combination with forced displacement constraints) also gives good control of unloading so plastic strain can more easily be seen.