Missing categorical variables (factors) produce error in Boost predict function after sparse.model.matrix transformation in R

Question

I am running an XGboost function in R. I prepare several variables by transforming first into factors (e.g. day of month). I then transform by using a sparse model matrix and