They probably calculated the original chances assuming independent random variables for answers, and the truth is that those variables are dependent and highly correlated because they are twins.

Both the articles and the original out of thin air calculations are completely made up, but it's pretty conceivable that the cheating program had a mistaken assumption of variable independence, which is clearly false in this case.