I didn’t even really start to think about Reuters and some of the others having these fluctuating samples, but I think that does explain a lot . . . except IA.
Iowa is unique primary. Historically about 5% of voters participate in caucus. Low budget candidates can win Iowa with lot of leg work. Example: Santorum in 2012. Iowa has not been a good indicator of national primaries.
In my opinion, summary of first 4 primaries is a much better indicator. (IA, NH, SC, NV)
Then comes super Tuesday (12 states). Results here pretty much will determine the eventual winner.