One of the purposes of an LGR is to confiscate an individuals firearms.
If they know the PAL# they want to confiscate from then:
a) they look at the list of ReferenceNumber buyer transactions for the PALer they're interested in, and find out what stores they bought from (seller PAL#), and go to the store and look up by ReferenceNumber/Date the make/model/serial of the non-restricted firearms he bought. That information is there for 20 years.
b) if they bought from private individuals, then they use big data to find that seller PAL#'s web forum userids, and use their copies of forum sell information to find make/model (but not serial) of the non-restricted firearms that were sold to him. If the seller is selling one firearm that within two weeks, this works pretty well. It's a bit hit and miss if they didn't sell/buy, or if they instead sold it to another getting another ReferenceNumber for the second sale, or if they're selling multiple firearms within the two weeks.
So, it's a slower LGR for both (a) and (b), but big data could make the latter (b) almost as quick but less accurate as the old LGR.
Variations on this work if a firearm is changed from non-restricted to prohibited.
(a) you use import stats to find the stores that sold it, and run through the store's CFO logs finding everyone who bought that make/model, and get the serial numbers and PAL#, and go and confiscate.
(b) you use big data to find the make/model, and get the seller PAL#, then use the ReferenceNumber database to find the buyer PAL, and go confiscate.
Both of these are a lot of work and a big waste of money, but doable.