You always get different data depending on where you pull it from. You say only 50% of games have fallen within 8.5 points, I've read several things stating a 6 point teaser changes the outcome around 65% of the time. I agree it's confusing.
I think the phrase "Efficiency of a Market" not only pertains to how close games actually fall to the predicted point spread, but how even the distribution is between favs and dogs, over and unders .
Favs/Dogs are very close to 50/50 over the years, as are overs/unders.
So by a market being efficient, it's stating that betting into a closer will give you around a 50% chance of winning your bet. By beating that closer by X amount, you have increased your chance from 50% to whatever the value is of each of those numbers you are grabbing.
In addition, I would guess the nature of football and the way points are scored in clusters has a lot to do with why you see several scores no where close to the closing line. If a game total is 42, it would not be at all unusual to see 1/3 of your expected total in a very small time frame. You would never see 67 points scored within a few minutes of an NBA game.
I'm just as surprised as everyone else that the average margin from the closing number has not improved at all. Where's Comptrbob? I'm sure he'd put us all in our places.