From the paper, the radius coverage metric is inefficient. Moved that remark here instead of losing it from the paper.