When querying on ratings with a constituent query, each criterion is evaluated compared with the constituent, not each of their ratings. This means that if a constituent has 3 different ratings, each of which meets a different criterion, that constituent still meets the criteria of the query. This means they will show up in the results. 
For example, consider a constituent who has multiple prospect ratings:
Ratings record 1: Prospect Rating date: 1/1/2012, Blackbaud Analytics Custom Modeling Service Target Gift Dollar Range: $25,001-$50,000)
Ratings record 2: Prospect Rating date equals 7/1/2012, Blackbaud Analytics Custom Modeling Service Target Gift Dollar Range: $50,001-$100,000

A query on the following criteria will include this example constituent: 
Prospect Rating Date equals 7/1/2012
AND Blackbaud Analytics Custom Modeling Service Target Gift Dollar Range $25,001-$50,000

This will occur because the constituent has a rating that meets each individual criterion, even though the same rating doesn't meet both.

Note: When querying on ratings, you will need to manually type in each of the rating descriptions, and they will need to match exactly how they appear in the system. For example, $50,001-$100,000 rather than 50001-100000. 

For a complete list of categories, refer to What are the categories for Blackbaud Analytics scores . For a complete list of descriptions within these categories, refer to What are the possible categories for Blackbaud Analytics scores.

If your organization is not using Blackbaud's Analytics services, ensure that your filters match the conventions of the data input on the records. For example, to locate constituents who have a Wealth Engine Estimated Giving Capacity, set up the query as follows: 
Prospects > Ratings > Specific Sources > Wealth Engine Estimated Giving Capacity 1 not blank 
OR Prospects > Ratings > Specific Sources > Wealth Engine Estimated Giving Capacity 2 not blank
OR Prospects > Ratings > Specific Sources > Wealth Engine Estimated Giving Capacity 3 not blank