Keep the result dataframes of the CONSTRUCT query around in the Spark session(now they get de-allocated), but do not save them persistently as in CREATE GRAPH
Temporarily add the meta-data of the CONSTRUCT query to the catalog under graph name X (the new catalog should not be stored in disk)