Well you have to remember birthday paradox...
In a room of about 20 people you have a 50/50 chance of having the same birthday as someone else in the room. What they're doing here is taking one sample and comparing it to hundreds or thousands of men.
I've been lead to believe that birthday paradox also applies to DNA too... So if you take a sample size of thousands or so presumably unrelated people you might find people who match in such a way that it would be acceptable evidence in a court of law.
Someone smarter than me can do the math but 1:1,000,000 chance vs 3,000 Samples ... How many are likely to come up close enough to prosecute?