I think you're missing Taylor's point: Launch vehicles are so damn expensive because they need to be extremely reliable. They need to be extremely reliable because every failure destroys the payload. The payloads are very expensive.

Assume payload cost is fixed. You can either keep trying to increase reliability, making the vehicles ever-more-expensive; or you can reduce the consequences of failure by finding a way to allow the payloads to survive, to fly another day. If you do that, you can use less-reliable, and cheaper, vehicles.

Somewhere in his paper he offers a comparison to commercial aviation. If a rocket engine explodes, the vehicle explodes and the payload is destroyed. But if one engine on an airliner explodes, you can "just shut off the fuel valve", as he put it, and the plane can still land--its payload of passengers or cargo can survive the failure. (I assume he's exaggerating a bit, because I can imagine an aircraft engine explosion taking out the wing; but I can also imagine that it won't always cause structural damage, either.)

The implication if that airliners were designed like rockets, to asymptotically approach 100% reliability of all components, air travel would cost as much as space travel. Or maybe he's saying that we should design space launch systems on the assumption that they'll fail every once in a while, and find ways to deal with it.

Conversely, I think you're advocating the economy of scale when you suggest NASA pay for a lot of launches of water. Taylor addresses that too, and points out that it's a chicken-and-the-egg problem. There's no incentive to build infrastructure for mass production until the demand is there, and the demand won't be there until economies of scale reduce the cost.

I still think he's found one piece of string we could pull on to make the whole Gordian knot unravel.