If a stereotype already exists, it is more easy to keep believing in the existing one (confirmation bias) than to invent a new one.
Except as you’ve just pointed out:
Some stereotypes don’t remain stable across generations.
Identifying a mechanism pushing towards outcome X is not inconsistent with observing that sometimes the outcome not-X happens.
Except as you’ve just pointed out:
Identifying a mechanism pushing towards outcome X is not inconsistent with observing that sometimes the outcome not-X happens.