Instead of returning an error if a model cannot handle static, historical exogenous or future variables, why not ignore the extra data and run regardless? Let the user know the variables weren't used. Right now, we have to read through the code itself (or the papers) to learn which variables we can use.