At the end of the day, these grants are so small because that's how much NSF is willing to give to people in pure math. NSF's math directorate only has so much money, and so they need to decide how to split this among all applications: fund a small number of larger grants, or fund a larger number of smaller grants. The decision NSF has made (correctly, in my mind) is to go with the latter approach. That means that a reasonable cross-section of working mathematicians (in pure areas) have a chance at getting a small amount of money to support their research, even if it's only a small amount. And apparently they find that useful, because they apply to these programs.
This answer is so matter of fact because your original question does not lay out an alternative. If what you really wanted to ask is why the grants are not larger, then that's because of the answer above, plus the fact that NSF simply doesn't give the math directorate more money. If what you really wanted to ask is why people apply for comparatively puny amounts of money, then that's because it's better than having nothing at all. Recall that if you're paid only 9 months per year, then even getting a tenth month paid from a grant equates to an 11% pay raise. Whether or not the grant is sufficient to really do substantial research becomes secondary in that case.