I think the improvement is solely due to the changing of the clusterd index to be the date key as the non clustered index on date probably would not have been used due to select * lookup cost being too great for very large amount of rows.

So I think the improvement is due to the clustered index key all by itself and not to do with the partitioning concept.

You might find that an unpartitioned table would have been faster if clustered on the date key due to better use of paralellism.