From a fragmentation point of view, there's no difference. If you run a shrinkfile after either dropping or truncating those tables, you will have fragmentation, and the same fragmentation either way.
TRUNCATE TABLE will free the space consumed, but DROP TABLE will also drop the definitions. If you aren't using those tables after you're done, then drop them. If you will receive updates to the raw file and will perform this work down the road, then truncate may a better choice than drop, so you can re-use the tables.
How big are the tables and how long does the import take? On the surface, it looks like you should create those #tables in tempdb, and just let them go out of scope when you're done with them.