The numbers don't lie. If you take a four year trailing average of recruiting rankings, turn those numbers into an ordered list, you find that about 70% of the time in head to head matchups, the team with the higher average wins. That number jumps to about 90% in BCS title games.
Don't confuse rankings/polls with quality of a team. If polls were correct, the higher ranked team should beat a lower ranked team with consistency. That means hypothetically the 8th best team could have 7 losses (If they played the top 7 teams), but there is no way the current polls would allow for that. The end result is a polling system mixed with good teams and those teams with the best records.
With a 70% prediction rate in individual games it should also be noted that about 70% of teams finish within one game of seasonal predictions. In the SEC, for instance, the largest over performer is Mizzou (+5 games over talent average predictions), followed by Vandy (+3) and then MSU (+2). The largest under performer is UF (-5) followed by UT (-2). The rest of the teams (65%) were within 1 game of predictions. 4 team's SEC record were predicted precisely by talent averages. This was a down year in the SEC as historically the prediction rates are about 80%.
By the way, Auburn ' s record was exactly what talent predicted.
Here is a chart that illustrates this year's predictions based on talent averages.
On my phone so sorry for difficulty with chart.