West Virginia attorney general Patrick Morrissey won Tuesday’s West Virginia GOP Senate primary with 35 percent of the vote, while Congressman Evan Jenkins came in second at 29 percent. Don Blankenship, a bigoted coal baron who spent time in jail for wrongdoing connected to a mining disaster that killed 29 workers, came in third at 20 percent.

In the end, Jenkins’s strength in his home district was not enough to overcome the attorney general’s advantage in the rest of the state. Jenkins was also hurt by Democrats who spent nearly $2 million to defeat him (while laying off the other candidates).

What happened to the much-feared “Blankenship surge”? On Saturday night, Politico reported that “four Republicans said they’d reviewed polling conducted in recent days showing Blankenship … moving narrowly ahead of his more mainstream GOP rivals.” On Monday morning, Donald Trump tweeted that Blankenship couldn’t win in November and Republicans should vote for Morrissey or Jenkins.

A few hours after Trump’s tweet, I reported the weekend internal poll numbers that both Blankenship’s rivals were circulating showing Blankenship narrowly ahead, with one poll pegging Blankenship’s support at 31 percent and the other showing Blankenship’s support at 28 percent--significantly higher than the 20 percent Blankenship ended up with on election night.

National Journal’s politics editor Josh Kraushaar reported Tuesday night that a lot of Republicans “really thought Blankenship would win it. Genuine surprise--and relief that GOP dodged a bullet.” But polling analyst Nate Silver suggests that the “internal ‘polls’ supposedly showing Blankenship surging may have been bullshit all along and driven a false narrative about the election.”

My report included the disclaimer that primary campaign polls are “ often unreliable,” without elaborating on the reasons that’s true, especially for internal polls: As readers of political news recognize, political campaigns have political agendas. But what was noteworthy here was that “ both rival campaigns”--not just one campaign or some outside group--showed a Blankenship surge (that’s a detail that hadn’t been reported by Politico). Why would they both conspire to make that up?

If Jenkins and Morrissey thought Blankenship was trailing by double digits, Jenkins and Morrissey ought to have continued to beat each other up and ignored Blankenship, as they did in the Tuesday, May 1 Fox News debate. But both Morrissey and Jenkins changed strategies in the final days of the campaign. As a Jenkins campaign official told me Wednesday: “Morrissey turned his guns in the press against Don.” Morrissey held press conferences going after Blankenship and sent a letter to the convict’s probation officer. “On TV,” the Jenkins official continued, “we shifted more of our weight at the very last minute to positive, rather than exclusively negative ads on Morrissey.... We thought we had more to gain my closing the campaign on a mostly positive note."

Since Tuesday night, I’ve learned of another factor that contributed to Republican anxiety about Blankenship: A Republican pollster unconnected to either campaign conducted a poll on Thursday and Friday that found Blankenship in the lead by “low double digits” (though there was a steep decline between Thursday and Friday). In the middle of last week, following the May 1 Fox News debate, the Jenkins campaign similarly found Blankenship briefly holding a similar lead. Over the weekend, both Jenkins and Morrissey internal campaign polls showed Blankenship with very slim leads.

So you can believe that Jenkins’s campaign and Morrissey’s campaign and an outside Republican invented polls showing a Blankenship surge, those fake numbers made their way to the White House and McConnell-world, and Jenkins and Morrissey decided to lay off each other and attack Blankenship even though they truly believed they were neck-and-neck and Blankenship was far behind.

Or you can conclude that the Jenkins and Morrissey campaigns and other Republicans genuinely feared a Blankenship victory based on their polling data, and the disparity between Blankenship’s standing in the internal polls and the election results is best explained by the fact that the polls were simply inaccurate (but not fabricated) and/or that Trump's intervention had some impact.

Trump’s anti-Blankenship tweet received wall-to-wall coverage in West Virginia. Only 136,000 people voted in the GOP Senate primary in a state of 1.8 million people. It didn’t require moving a lot of voters to swing the race several points. But one campaign official suggests that the automated polls, the method used by both campaigns internally, may have been biased in Blankenship’s favor. “Either there was a little blip that was exaggerated, or there was something in the methodology of the IVR [instant voice response] polling that inflated his numbers,” the campaign official says. “The numbers that you wrote about are the same numbers that somebody [already had] passed along to the White House and to McConnell world that launched the freakout,” a Republican strategist following the race told me Wednesday.” Internal polls showing a Blankenship surge (accurate or not at the time they were conducted) explained why Republicans had been behaving the way they were.