Better teams don't necessarily perform better. Just think of a team as generating random numbers from a distribution or a range where some numbers are more likely than others (points per possession, points allowed per possession, whatever). Based on what the team has done in the past games, you rate them by which team is expected to produce a higher point differential or something similar. So, a truly inferior team could win a single or just a few consecutive games (Butler) just by chance. You could roll two sixes on your first roll, even though this is the least likely outcome just like a team could win or lose one or two or three games but still be expected to be better than it seems. This doesn't even consider that those losses could be against good teams and having really good game numbers like turning it over twice.
If some teams are getting ranked by what they've done (SU: #2 in official polls, 1 or 2 in pure win-loss rankings but as low as 10 in predictive rankings) but others seem to be getting ranked by predictive systems (UMass: #22/23 in official polls, top ten in pure win-loss but low 20s in predictive rankings) that doesn't seem right. Then there's Duke that's 14 and 15 in Sagarin and Pomeroy, in the mid-20s and low 30s in pure win-loss but somehow gets to number 7 in the "official" poll. I guess they want a top ten match up when they come to the Dome.
