Sagarin takes into account margin of victory and fully builds in home/road adjustment. Similar to the systems the folks in Vegas use so I prefer it over RPI. But RPI seems to be the record of reference when people talk strength of teams.
I think RPI is record of reference often because the selection committee uses it (by looking at a team record vs. RPI top 50, top 100, etc.) when determining at-large bids.
I think Sagarin typically gives you a better picture, especially early in the season. The sample size is still small enough, and scheduling oddities play a factor. Also, Sagarin's ratings early in the year include a base rating of a team (a preseason rating) -- those are still in play at this date. Sagarin rates us 87 with his ELO_Chess, which doesn't include point margin -- much closer to the RPI formula, while we rate 216 in the Predictor (which factors points heavily). His overall rating is an average of those two pieces.
The two formulas tend to match up better the more data you put into them.