Following the links from the OP, I see the first researcher states there are fewer than 2 NINA ads per decade in the Brooklyn Eagle, while following the example of the search I counted more than 8 ads in Brooklyn Eagle from the 1870ies before giving up the point as discredited.

I wonder if the big reason for these completely different results in the underlying counting experiment is that the later research was done after ocr had improved so much that the searches had very different samples of the papers.