You can manually create a data table with a column having the proportion you want with the desired run size. In Custom Design, select Add Factor -> Covariate, and choose that column. Now you can add the rest of your factors, model terms, etc..., and keep the number of runs the same when creating the design.
Thanks Ryan. I've not used covariates, so I will have to tinker with it some.
I did have another idea. Create a dummy 2-level factor and disallow combinations of factor=on & dummy=on
This would bias the distribution of factor approaching .33 “on” and .66 “off”, but at the expense of an additional factor, driving up the # runs. This only works if I require the dummy factor main & interactions, otherwise it doesn’t strongly drive a bias of state.