Article dans une revue: The final step in the proof of Proposition 1 (p.311) of Mukerji and Tallon (2003) may not hold in general because $varepsilon>0$ in the proof cannot be chosen independently of $w,z$. We point out by a counterexample that the axioms they impose are too weak for Proposition 1. We introduce a modified set of axioms and re-establish the proposition