I would argue that any time accuracy (or any feature) is expected, then a lack of said feature does matter. You never know when someone somewhere could be relying on the fact that they can get within 5 yards.
Look up "circular error probable" (CEP). When they say "accurate to x meters", they might well mean that 50% of munitions land in an x meter circle --- the other 50%? anywhere (maybe gaussian distro). So all this talk of x meter accuracy is quite misleading.
That may be true, but (assuming a normal distribution [1]) that would also imply that 99% of shots land within 2.6x meters (assuming I didn't mess up my arithmetic).
Whoever is doing the aiming gets a big red circle that encloses the margin of error. If you are off by 5 meters, that big red circle is in the wrong place. The pilot expects a 1% chance of error, in reality it might be 4% (i.e., 4x as many screwups).
[1] Of course it isn't a normal distribution, but this only changes the programming details.
