Here are some observations on querying and materializing the triples received from a Postgres DB (these all use the same graph and the same WHERE clause in the body):
Query (select) output was pretty quick (I didn't note the time but it was pretty fast)
CONSTRUCT output was much slower but still ok (3 minutes)
Thanks Andrea. Were any of these run with reasoning? How many triples were returned from these queries? How many triples were in the database before the INSERT?