By specifying before the window that Model B equals 2-to-12-hour release the coalition prevents the temptation to reinterpret results after seeing the data