The references he gives in that section suggest that it boils down to a convenient way to try to gauge poison by contrasting sugars to bitterer phytochemicals. Plants with sugar may be less nutritious on average, but that's just counting calories/macronutrients, not taking into account poisonous substances.