This isn't surprising at all since you would expect a lower percentage of new users to have high rep since, by definition, they have had less time to build rep.
They also are less likely to ask really good questions or give really good answers since...
They are learning how the site works.
There is by now a very small pool of certified experts who don't already use the site.
Due to (2) above most new users are likely at "apprentice" or "journeyman" level in their fields.
As the site's popularity grows, we are likely to have more "hit and run" users that join to ask a single question, get their answer, and leave.
Also I think judging a "good" user based on whether they have 100+ rep is a fallacy in itself. The old BAD users have just had enough time to get 20–30 sympathy upvotes on their high volume of terrible questions to pass your test.
EDIT
Here's a new version of your query that I think is a better indicator of "good" users.
It shows your original, plus total votes per user, and (this is the part I think is most relevant) number of users who average 10 votes/month for the life of their account. That number has been under 10% going back to users who joined in 12/2008, so as far as how engaged people are in the site, it looks like we are pretty consistent.