I have read the statement that QED has four counterterms to cancel divergences. However, I have learnt that there are only three counterterms (vertex, electron propagator, photon propagator), which is consistent with this PDF$^1$ (42KB).
But then, the last page of this PDF$^2$ (229KB) states that there are five kinds of diverging skeleton diagrams in QED, one of which vanishes (the triangle).
Is there a fourth counterterm or did the original statement actually refer not to counterterms, but to diverging diagrams?
$^1$ "QED Feynman Rules in the Counterterm Perturbation Theory" by Vadim Kaplunovsky (University of Texas)
$^2$ "Renormalization in QED" (Imperial College London)