I agree that you shouldn't be forced to use all of the units in a component. I've seen plenty of designs that use dual op-amps everywhere, and some of the chips used only one of the two units. And that was for layout reasons, mostly. "But you wasted an op-amp!" goes the cry, but it costs more money to have to buy both the single and the dual op-amp chip, and it costs more to have that extra slot in the pick-and-place machine.
My reasons for requiring that an unused gate get put on the schematic are two-fold. One is as @dchisholm notes, sometimes unused inputs need to be terminated or both things happen. The second is just so the human technician who's looking at a board can go, "Oh, U26, only A is used, B is not." It's made explicit.
I find the arguments against "littering" a schematic with "dozens of supply symbols and/or bypasses" tiresome. We need to be explicit, so that there are as few mistakes made as a result of unstated assumptions as possible. (My earlier comment about "tribal knowledge" regarding bypass caps notwithstanding, of course.) I want my techs to know that +5VA and +5VD are separate nets. I want my technicians to know which bypass caps are associated with which chips. Schematics convey intent, and if that means text blocks, then add them. If that means circles and arrows, add them.
Remember that the engineering notebooks aren't archived with a design, and sometimes the original engineer isn't around to explain what was done -- and why.