Bug Fix: If we reach a denormalised result and it's still not small enough, return zero as we're not going to find a better value.