Begich up 17% with early voters, down 6% with those certain to vote, and somehow down 5% overall? Yeah, alrght. Lol.
It's Rasmussen-less Rasmussen, so take any of those numbers with a grain of salt.
However, there's not enough early vote out really to make that much of a material difference. You're probably talking about a sample size of 60 or so with a margin of error of roughly +/- 10%.
Early voting numbers in any poll should be taken with a grain of salt until there's enough of voters.
A Colorado Early Vote number where 63% of the 2010 vote has already voted would be much more reliable and statistically significant than an Alaska poll where there's as many as 9 out of 10 folks haven't voted yet.