I know they are considered the 'fisher price' of carbs, but if I was going to go downdraught, I'd run a 350 or (on a moderately modded a15) even a 500 holley. Avoid the holley 320 like the plague, they are like the 'emissions' spec weber sidedraughts - too much changed to be easy to get great performance out of.

The 'problem' with the factory stock intakes, is that the firing order is 1342, or 21 34, which means two intake events from the front pair of cylinders one after teh other. With enough cam duration,there's actually overlap of the intake events and one will try to rob the other of flow, one right at the end, one right at the beginning. This is a big issue on mini engines (who have a single siamese intake port for each pair of cylinders) and they even have 'scatter pattern' cams for race stuff with different intake cam lobes for 1 and 4 , and 2 and 3, one set opening and closing a little earlier than ideal, the other set a little later than ideal. Although both are then 'wrong' for the ideal cam timing for the engine, because there is less 'stealing' of intake flow from that shared port, the overall power is increased, and the engine has a wider powerband too.

A similar thing would happen with a stock factory manifold, but the stock cam is quite mild in duration so there's no significant overlap in each cylinder pair. But move on to bigger cams, adn it's an issue. And having a big runner means that whole volume of air/fuel will supply that cylinder. Then, due to 1 and 4 being the 'outboard' runners and 2 and 3 being 'inboard, they each slightly preferentially draw from the barrel directly above them, and since 1 and 4 intake cycles are a full 360 apart, there's no overlap.

So basically that 'full runner' setup is a winner on many fronts. What would be interesting to try is to actually mod them and fit a dividing wall so 1 and 4 could 'only' draw from the outer barrel, and 2 and 3 from the inner, full separation. That ought to make for some interesting mid range torque. At least it'd be worth experimenting with.