The current implementation of the reputation system is not about someone's reputation, but in fact just a representation of how much money they made and the amount of money a post made has almost nothing to do with the quality.
Current problem
Right now we CANNOT rely on up-votes and down-votes received (when someone is the author of a post) to accurately estimate that user's reputation, simply because people mainly vote on high earning posts just for the curation rewards. It's been a huge problem since the start!
Temporary solution
A reputation system has to accurately estimate how much you can trust a person to classify content. A good start is to verify content the user has voted on. Since all data on the blockchain is public (except upcoming private messages), you can check how many up-votes and down-votes each post has. So if you filter all posts to "voted on by user", you can calculate the up-vote to down-vote ratio. On top of that your reputation should also increase when you downvote a post that receives more downvotes. This should work the same way as curation rewards. The earlier you spot bad content, the higher your reputation increase should be!
Long-term solutions
I described in detail in one of my earlier posts (SteemIt voting algorithm problems and solutions from a programmer's perspective) how to solve the main problems we have right now. I highly recommend to read it if you are somewhat technical. Until these algorithms have been fixed there is no way we can see any correlation between how much money content made and it's quality!