To really answer the retrieval question, there are some tests we could do. The problem is isolating variables. For instance, we could isolate a large number of Normal retrievals for an item and compare different retrieval stats to different amounts of actual item obtained. But, that makes the assumptions that no other stat affects the number of items, the wind doesn't affect the number, and the sector doesn't affect the number. We can correct for some of these, but fixing all of them will result in our sample size dwindling to nothing. And then there's the question of what item to use - if we use a common item that's specific to a single sector or pair of sectors, like Bamboo Weave, we limit the range of Retrieval values in our analysis. The vast majority will be 70, and the rest will be near 70. Few, if any, people will send an airship with 100+ retrieval into SC01 or SC02 - there's simply no reason to do it. If we choose an item that's distributed among many sectors, like Birch Branches or Water Crystals, we run the risk of making invalid assumptions if the sector rating affects the retrieval amounts (like, if higher rated sectors require more retrieval for the same effect).

Regarding sector ratings and retrieval ratings, have you actually done that analysis? When I did, I found it rather inconclusive. But, it's a hard test to do, because higher retrieval ratings often also mean higher favor ratings, and that's what I believe actually affect the sector ratings.