It just seems really weird and un-natural to me, to waste 40*24 pixels of an image in order to center a turret, is all.
Makes it easier to draw barrels as another layer in the same PS file, I'd imagine. It's also easier to specify the weapon file manually - one less set of coordinates to worry about. The only "waste" here is of video memory, and it's negligible. It'd take 100+ different large-ish weapons to take up 1 meg of video memory (note that trimming barrels down would not take it down to 0). For comparison, a single Onslaught takes up about that much.
If the quad's trimmed down to the actual size somehow, no biggie, but if not, it's wasting a lot of texel operations, too.
It's not trimmed down - those texel operations are likely cheaper than computing an extra pair of coordinates on the CPU, though neither is much to worry about.
This is exactly the kind of optimization that's just not worth doing, until it becomes a clear bottleneck. Actually, pretty much all optimizations fall in that category. This one is just so minor that I don't see it ever being worth it, though - never mind that it's not a clear optimization, but a tradeoff - and introduces added complexity in a few places.