This is like one of those puzzles where you're supposed to find all the mistakes, like the tire on a car is a pizza. Your assumption, expectation and math are all wrong.
The article doesn't claim exit polls are useless - they have limited use and the author explains what they are (e.g. trends for talking heads such as "more black voters turned out than expected"). I wouldn't expect an even outcome of outperforming exit polls because it's 
not random. It's not coin tosses, it's supporters of candidates, and the candidates are the same across different states. So if Hillary supporters tend not to participate in exit polls, I'm not all that surprised if that trend continues across different states.
The one thing you said that's true: "I didn't do the math"
I did do the math 
on Wolfram Alpha and the odds of 26 coin tosses coming out with 24 heads and 2 tails is
4.84288 X 10(power of -6) or 1 in 206,489.
Not quite 1 in 77,000,000,000