What I did not see in the report is if this was just a sample of data, or if they had the entire voter datasets for the states in question.
I think the search criteria was laid out pretty clearly.
1. Equivalence of first, middle initial, and last names
2. Equivalence of birthdays
3. Equivalence of partial SS#
Those not counted would include:
1. If someone used a nickname for their first name, it was probably not counted.
2. If someone omitted their middle initial, it was probably not counted.
3. Non-citizens were not included.
4. Deceased citizens were not included.
5. Phony individuals (dogs, cats, etc)
etc., etc.
This is the tip of the iceberg. It states later on in the article that if these results were projected to all 50 states with only this search criteria, the number would be 45,000 double votes.