Without having read the article, this makes some amount of sense. Roughly speaking, based on our behavior everyone of us has a probability of catching the virus, and (assuming we did catch it), a number of people we will transmit the virus to.
Multiply those two values together, and you get an estimate for how many people each individual would infect in the long run. One might think of this as a "personal R value", because averaging these values over a whole population would result in something similar to the "R value" we're currently talking about.
Obviously, this value grows if either of the input values increases, and is highest for reckless people with many contacts and lowest for a cautious recluse.
Now, if you could pick just one person from a population and set its "personal R value" to zero, which one would that be? If you pick "sweet grandma" because she didn't do anything wrong and deserves it, you're helping her - but if her "personal R value" is something like 0.1 (low probability of catching the virus, few people to transmit it to), while there's someone else running around with a value of 20, taking grandma out of the equation won't do much as far as the overall pandemic is concerned, while taking out the potential superspreader does.
That being said, I don't think we can look at this as just a number's game. While vaccinating superspreaders first would be the best way to quickly stop the pandemic, it is also the most improbable way out of the situation we're in. After all, if these people would be willing to help, they could just change their behavior and stop being superspreaders in the first place.