Wenjun Li recently made us aware of a mistake in this paper. At the end of Page 9, in the discussion on covering points with lines, we claim a running time, for kernelization, of O(kn). However, this does not follow from the arguments presented and the correct running time, for kernelization, is O(kn log k), yielding an overall running time of O(k^{2k+2} + kn log k).