Data Discovery relies on many objects found in your database. Views, Foreign Key Constraints, etc. Sometimes the objects in your database can be incorrect. Yea, really!
Since DSF learns from your database it can learn bad things. The best way to solve this problem is to create a new DSF database and select the None option for discovery.
This will create a blank DSF database structure. You can then slowly teach DSF the correct way to access your data model.
To accomplish this, you use the Using SQL Queries option found on the Data Discovery tab.
Collect SQL Queries from your current reporting tools or co-workers, that you know correctly access the data. Paste the queries into DSF to discover. The more queries you feed DSF the more it will learn.
More details on discovering data with SQL Queries can be found here.
If you are still having data discovery issues you can always contact us for help.