Hi, good question; these datasets are indeed meant to be fully aggregated, thus "LCI" processes (in openLCA nomenclature), but some contain input products which means that they contain incomplete supply chains that are not followed up to the elementary flows. This can be result of a calculation in GaBi with looped systems, which is terminated before a loop is fully converged. You see the same effect in the older ELCD database.
So, you are right with the interpretation, but the PEF database is not perfectly providing fully aggregated datasets, for these cases.