I had the exact same question and I think the news article didn't do a good job to answer it. But the paper does claim that his control was terrible, even though not clear why because he didn't seem to document it properly.

> After addressing problems with Blodgett's analysis and control experiment, we found results indicating a twist of only about 7°, but a much larger “hysteresis” effect (∼19–23°) where perceived direction was drawn to the horizon regardless of dazzle. This effect combined both constructively and destructively with “twist”, depending on the direction of the target ship.

Then specifically this part is very suspicious, zero error is unlikely in any circumstance:

> A uniform white and black background were assumed to produce zero errors. In the case of white, Blodgett (1919) claimed this was borne out by experiment, though no details (e.g., number of trials or observers) were provided. Furthermore, there is no evidence (or even a suggestion) that the black condition was run.

> Results were also reported for ships painted uniform black and uniform grey (e.g., see Table 2 in S2). These conditions were not mentioned by Blodgett until his results, and it is not clear under what seascape and weather conditions these controls were run, how many trials were performed, what the true directions were, or which or how many participants took part. In fact, there is nothing in Blodgett's report to suggest that these conditions were performed by any of the six participants mentioned above.

But I think the researchers' point is more about how Blodgett may not have accounted for how much that effect was influencing perception

I suspect the effect is similar to why zebras look the way they do. Patterns and perceptions of them aren’t always super intuitive