The values stored in the returned marginal model are just from the final run of the simulation (not very useful). To save results from all 5000 trials, you need to modify the simulation definition with the Mimi.addSave!
function described above, which will save the results from all trials to a file (and then you will have to read the files in and perform desired averages/medians yourself).
And yes, that’s what I meant by the pulse size is one million for ten years-- one million in each year for ten years.