This article might be interesting if it said one single thing about the polling methodology or who was represented in the sample of 1,000 people, of which only a sample of 500 was selected for RCV. What can we really learn about a national election from a sample of 500 people about whom we know absolutely nothing except that they’re “voters?”
The theory about RCV eliminating a Keynesian Beauty Contest (KBC) is also extremely dubious in the context of a primary. In a primary, voters are still going to vote for the candidate they believe will be able to prevail in the general election. In other words, their first choice is still for the person they think has the best chance of winning.
In that light, the closest thing to real results that you were able to obtain, where you did not juggle any data and you ranked the top 5 after polling people for their top 5, comes as no surprise when it shows that there is no meaningful difference between RCV and KBC in the final rank of the candidates. As of the date of this poll (which you neglect to mention is two weeks old as of the publication of this article), whoever these voters were (we have no idea if they’re all Democrats, or Democrats and Independents, or random Americans, or what) probably believed that Joe Biden had the best chance of defeating the GOP nominee.
RCV in the context of a primary surrounded by horserace political media coverage is likely of no value in terms of bringing political outcomes closer in line with democratic political will. This would be far more interesting in general elections with robust third-party representation, or in nonpartisan blanket primaries, where we have already seen it put to good use.