Alan Schwenk points out that the assertion in the last line of Section 2 is not correct, and that the construction gives slightly smaller counterexamples than the assertion implies (of order as small as 205, as opposed to 246). He also suggests a modification of the construction that gets the order down to 106.