I'm currently looking into these quackle assumptions.
I don't think the values have a whole lot to do with word knowledge. The way I see it, a person needs to learn every single word in the dictionary before they even have a chance to play the game properly. But beyond that, things tend to average out, and if you study high probability words, you should actually prefer to keep a better leave over a worse leave for more points. So I tend to disagree with your assumption.
If you know less words, you should keep leaves that you are comfortable with, which will hopefully be good leaves.
The values, if they are wrong, have to do with the assumption that the opponent won't play defensively. I have a lot of trouble grasping how much this actually matters, but from what I can tell, there is a little bit of risk in giving away that you have good letters. This might be a minor mistake of Quackle (not going to call it a major mistake quite yet).
Word knowledge is irrelevant though. Just learn all the words, period. I know this sounds ridiculous, but at least the top 10000 bingos you should know pretty comfortably if you wish to play competently. 90% of scrabble is word knowledge. Strategy is only effective if you actually know the words.