Correct me if I'm wrong (and I frequently am), but if the advanced stats are describing what Indinia has done this year against a beyond weak-sister SOS, would those same stats tell a different story if they had played a top 25 SOS? They played one of the weakest schedules in the nation, and I would think their statistics, while indicating a good team against below average opponents, would reflect that they are better than if they had played a tougher schedule? Admittedly, there are a gozillion folks here on Loyalty that digest and follow the stats more than I do, but I seem to recall that the stats most use are descriptive of past performance.
So when utilizing stats, it's important to know that the statistics of anything won't tell you what will happen, just the overall likelihood of something happening when you consider x, y, and z specific variables.
When it comes to statistical engines to analyze college football teams, there are two different major philosophies: There are the predictive based engines (these are the far more complicated ones) that try utilizing the player, team, performance, and game data to predict a score between two teams or how a team will perform throughout a season. This is difficult, there is a lot of error involved, and those that do it well still tend not to outperform Vegas on the whole (who are also doing it themselves, just better). Then there are the performance based models, that don't really care about who will win, but instead takes all game data and iteratively analyzes each team to discern how a team's performance compares to other teams. These engines are simpler than the predictive engines, but they do have their own complications, especially because 12 game data points in the grand scheme of things is quite small.
So in your question you are asking for a predictive solver around Indiana, however, what you really are asking is how does Indiana's performance this season compare to SEC 3 loss teams with much tougher SOS. And the performance based solvers will tell you that Indiana compares very favorably. And that makes sense, they've blown out pretty much all the average to good competition they've faced while only losing to a Top 5 team on the road, whereas the 3 loss SEC teams have multiple losses some to .500 teams, teams of similar strength Indiana has destroyed. So Indiana gets the performance engines seal of approval.
However, the predictive engines, based on their inputs might still actually favor those 3 loss SEC teams over Indiana as they take far more into account than just on field performance. It's why if Indiana was playing say Alabama in the 1st round of the playoffs, the line might favor Alabama. It's simply because the performance engines will say Indiana performed better this season relative to Alabama after accounting for schedule, however the underachieving Alabama would be predicted to beat the overachieving Indiana on a neutral field.
Does that help make sense of this a bit better? Oh and just as an aside, I am always always always in favor of Performance based engines when it comes to seeding. In my opinion, predictive based engines should be used for betting and prognostication purposes, otherwise you aren't rewarding teams who actually performed the best, you're rewarding teams with more talented rosters. You want to make the CFP, then win the games on your schedule. You lose 3 games on your schedule, and you should thank your lucky stars if you get in, because you aren't all that deserving of the spot