My guess on the Catalina is that, intuitively, the area of wing adjacent to the body is pretty well compromised by the slow (relatively) air arriving at it. So to free it from the fuselage gives the centre section a clear view of still air.
The reason for my interest in this thread is that as of about five or so ago, I read that Boeings Kray-hosted CFD could model wings and fuselages, but not both.
My design interest is mainly sailing boats. Our foils have to work both ways up, as it were. Keel mounting in cruising boats is often a grp moulded keel moulded with the hull, and filled with heavy stuff.
Racing yachts and cheap boats have a bolted on keel.
Over the years I have noticed that the moulded grp keels are filleted, bolt-ons are not. Reason being that grp would have to be massive if a hard point were introduced. Fillet distributes the strain more evenly.
So as a young man working at a British Shipbuilders towing tank facilty at the time of Dr Pieter VanOosanen and Ben Lexcen's remarkable "upside-down" and winged keel on Australia III, I got to thinking. At the same time, we were running some submarine tests and used foils to hold it submerged. I wondered if it might be a good idea to separate the keel from the hull entirely because, surely, the water in the region of the non-fillet couldn't be that helpful.
Fast forward again to when I read of Boeings' computer problem.
The New Zealand America's Cup boat was revealed with a 'dillet'. Reverse fillet, so that water arriving at the hull/keel join was encouraged in by the low pressure. Made sense to me, and seemed to work for them.
Back to 'planes. Fillet might add strength if it were integral with the body/wing skins, but I suspect that it's not really.
So why fillet, and why not dillet on the top side? Get the air speeded up a bit.