can't dispute the conclusion but I object to the argument.

Because, the argument doesn't apply to this model: "Probabilities are calculated from a random walk, except for large unlikely changes, or changes when the price is near zero."

Approximate models always have a range of applicability. Showing that the model gives absurd results outside the range of applicability for which it's intended doesn't mean much.