Wasn't this "source" discredited like years ago?
It's not discredited, it's just that it's easy for people to misinterpret it to say a lot of things it doesn't. "Liberal" or "conservative" measures are based on statistics of deviation from the norm in voting patterns, rather than derived from any specific ideological content of the votes themselves. The model assumes that Republicans are on average more conservative and Democrats more liberal. And so if a Democrat votes a lot like a Republican, he's relatively conservative. Ron Paul voted extremely differently from the average Democrat, even if sometimes for socially libertarian reasons where he was the only one to vote that way, so that contributes to him being placed as the most conservative.
This article by Sean Trende does a good job explaining the basics of the scores.
https://www.realclearpolitics.com/articles/2012/05/11/what_has_made_congress_more_polarized.html
