You said you're looking for something efficient - for such a relatively large data set, having the DB do the work should almost always be more efficient than reading the data into Perl and doing the work there.

If you're worried about getting the queries right, how about setting up some test cases? A few minimal sets of data for which you can figure out the desired output "by hand" for the various queries you want to run, and then work on the queries until they match the desired output.