Mike Cohn wrote an interesting post where he discusses he allows or even encourages to estimate with story points as large as 20, 40, and 100. He explains that they are useful when you need first and not necessarily precise estimate of the general size of a new project being considered.
His post generates an interesting discussion about the necessity to break requirements in smaller pieces, the difference between estimating and committing, playing Planning Poker during sprint planning or educating the stakeholders.