The other thing is this is only for process values, the mismatch information is still scattered in each of the device models. To me it would make sense to take those out of each model and store at the top level so you could easily enable/disable mismatch variation modelling by either importing a file with the random function or the file with all values zero