Democracy does not guarantee efficient government, it is correct. However, dictatorship does not guarantee efficient government either. Dictatorship can make decision in a much shorter time than democracy, but it is only the input side of efficient. To determine efficiency, I believe we need to consider effectiveness and quality as well.
Of course, dictatorship can make wrong decision and democracy can make wrong decision. I am not sure whether Prof. Cheung is going to receive money from his bet. Even I presume the probability for dictator to make wrong decisions is lower, it does not mean dictatorship is better in this sense. We need to think about the severity of those wrong decisions. For this, I bet democracy is more able to reverse wrong decisions and control their damage. In short, the expected damage of wrong decisions under dictatorship may (not must) still be greater even its probability to make wrong decisions is lower.
For me, transaction cost for different types of government is ranked as:
Good dictator<Good elected leader<Bad elected leader<Bad dictator
Democracy, therefore, will never achieve lowest transaction cost but have a lower uncertainty. Again, expected transaction cost for democracy may still be lower. I personally believe that democracy is adopted to limit uncertainty rather than achieve highest possible efficiency.