I don't have any piercing insights into why we saw those two features over the past two winters. As a starting point, I would go back to CPC's seasonal attribution page that I mentioned in a previous comment. For this past winter, the verified North Pacific high anomaly was a bit west of the anomaly in both the forecasts and the simulations that were forced by the observed sea surface temperatures (the simulations labeled as "AMIP"). This would suggest that chaotic weather variability may have been the culprit, and if so, then we would have to look for subseasonal phenomena (like the Madden-Julian Oscillation or upstream Rossby wave trains) for further attribution. But even if we can point to such phenomena, they would not be too relevant for seasonal predictions since they vary quite a bit from week to week. 

It's pretty much the same story for the previous winter. I will note that Siberia has a very wide range of chaotic weather variability ("weather noise"), which means they can have pretty big weather swings that are not evident in seasonal predictions. There is some debate about the possible roles of Arctic variability (like with sea ice) that perhaps our models do not capture well, but at this point, chaotic weather variability rooted entirely in the atmosphere is a difficult null hypothesis to reject.