I'm not sure how you can square anything you said in your last paragraph with anything I said about all models being wrong, and causal modeling being extremely limited.