Interesting, I think how the model runs makes a big difference and I plan to re-run this experiment with different models and different ways of running the model.