Short version:
Warning popups --> Quantity of reports over quality
Actual punishment --> Quality of reports over quantity
More detailed version:
There are two primary outcomes to consider with a system like this:
True positive (deserved punishment and received punishment)
False positive (didn't deserve punishment but received punishment)
The warning popups don't actually "harm" anybody; but they're important to provide instant feedback for people that might need help to reform as well as to spread the information that certain behaviour can result in punishment. That's why Riot's goal with warning popups is to maximize true positives.
To achieve that, Riot focuses (mostly) on the number of reports people receive.
When it comes to actual punishments, the idea is that you have to make sure that innocents don't get punished. That's why the goal there is to minimize false positives. This is where false reports become problematic, so you need a way to identify those. That's why actual punishment focuses more on the accuracy of reports. The actual way this is done is pretty complicated, but involves a lot of manual input and machine learning.