Well, actually there are multiple causes, but this is one of them. The launch escape system is very heavy. And it’s heavier than it needs to be because of the inherent inefficiency of the engines resulting from the cant outward (necessary to avoid blasting the capsule with the exhaust). Note that each opposed pair are fighting each other with the horizontal components of their thrust, contributing nothing whatsoever to the mission. This is called a cosine loss because the effective amount of vertical thrust is the total thrust times the cosine of the angle they’re canted at. Since the lost thrust is the sine of the angle, you need more thrust overall (and hence a heavier engine) to compensate, making a bad problem worse.

People have considered putting the escape motor underneath the capsule for this reason (I think that Mike Griffin even drew a napkin sketch of it–we looked at it in OSP as well), but that complicates jettisoning, since it goes between the capsule and the service module. That would mean that you’d have to carry it all the way to orbit on each mission, and then separate, jettison, and redock with the SM, which carries performance and safety risks in itself. Or if it goes under the service module, then the motor has to be a lot bigger, and then you have to do a CM/SM separation after motor burnout but before rotation for entry. So they stuck with the Apollo tractor configuration, in which the capsule is pulled away in an abort.

The other solution, which would give them a ton (actually, literally tons) of margin would be to get rid of the damned thing. It’s only there as a backup in case something goes wrong with the launch vehicle, and then only if specific things go wrong (for instance, a loss of thrust wouldn’t require it). The weight and design is driven by the extreme case in which the upper stage is exploding beneath you and you have to try to outrun the flying debris. This is an extremely unlikely failure mode, but politically, they have to have the system there, because no one wants to take the chance that they’ll have to testify before Congress that they killed astronauts because they didn’t have it. With it, the estimate is a one in five hundred chance of losing a crew. Without it, it’s much higher (though there are no doubt many astronauts who would accept the risk regardless, since they’re already doing so now on the Shuttle).

Also, as Jon Goff has pointed out in the past, they’re putting a lot of effort into safety during ascent, when this is actually one of the lesser hazards of a total lunar mission.

But that’s the way that politics drives a government space program, and why it is so horrifically expensive.

[Update a while later]

It just occurs to me that the other case where you need it is an on-pad, or shortly-after-liftoff abort, when there is insufficient altitude for safe chute deployment.

But the thing to keep in mind is that it made a lot more sense in Apollo, because in the early sixties, “our rockets always blew up.” The technology is much more mature now, and the failure modes for which it would be needed are much less likely, even in an expendable.