How does that follow? If you have a below-average diet and replace foods from it, you're likely to do better simply due to regression to the mean. You could replace a shitty diet with one with different foods but the same macronutrient distribution and likely end up improving things.