It is hard and actually not useful to get lost in the details of why someone's ranking changes between iterations. Basically, in the first iteration all players start with the same amount of points, so as a result of the first iteration the players will be ranked mostly according to the number of opponents they've faced and the win ratio against them. Then the second iteration starts with that ranking, and now players can be rewarded according also to who they beat, obviously beating someone from the top will get you more points than beating someone from the bottom, but who was at the top and who was at the bottom at the start of the second iteration depended mostly on their activity and win ratio and not at all on whom they beat, so perhaps the result of this second iteration is not quite reliable yet. On the third iteration you start from a ranking where win ratio matters, activity matters and who you beat matters. And so on, until the ranking is stable.