Checking for model consistency in optimal fingerprinting: a comment