Glue ETL Job with external connection to Redshift - filter then extract?

I have been attempting to ETL a data set from redshift with glue for data lake consumption. The redshift data set is very large and I only want to extract the last x days data each job run. When I setup the job the filter comes after the applymapping and before the resolve choice dataframes. When doing so the redshift query appears on the cluster as essentially a "Select *" . It seems the dataframe wants to load the entire redshift table into Glue then filter it, which is expensive and eventually fails. Is there a way to filter the datasource before the dataframe?