If you're trying to introduce new vocabulary by ensuring that it can be inferred from context, minimizing the number of new words is a reasonable strategy.
If you're trying to produce a more generalized grapholexical model, then you want the amount of new vocabulary to be too large to memorize one by one.
As always, you cannot judge the effectiveness of a method without an understanding of the goal.
So I would start by leading with the goal, and from there everything else should follow.
It comes down to what you're trying to teach.
If you're trying to introduce new vocabulary by ensuring that it can be inferred from context, minimizing the number of new words is a reasonable strategy.
If you're trying to produce a more generalized grapholexical model, then you want the amount of new vocabulary to be too large to memorize one by one.
As always, you cannot judge the effectiveness of a method without an understanding of the goal.
So I would start by leading with the goal, and from there everything else should follow.
Great advice. Sounds like this is something that could be explore in depth further.