(my focus and KYs are targeted at one 1mm population city with the city name/variants in all KYs, if that makes any difference)
Let's say your Ad Group has been running for six months, and for three of those months you had a very low Max CPC, and therefore a low Ad Position. Well, if you are looking at the date range of 'all time', those very low positions are averaged into your 'Avg. Pos'. And it'll seen way lower than what you'd expect currently.
In other words, the Avg. Pos. for 'all time' can be very different than Avg. Pos. for 'yesterday'.
So keep your eye on that date range. It is a sticky setting which stays where you put it. I make it a practice to glance at the date range every time I review stats.
AWA
That being said, buenavista makes an accurate observation.
I have seen the same thing across many keywords that are in various positions and that have various levels of competition. And yes, the natural tendancy is to overbid. But if one looks at their actual position on Google Search and the partner sites, a more accurate assessment of an ad's placement can be obtained. The average position stats give a good general indication of placement but are about as accurate as Google's traffic estimates.
I run a country specific campaign using exact match exclusively - so it is very easy to track this phenomenon. As far as date range, I keep mine on the current 'Today' stats only, and I see this discrepancy daily.
buenavista - Trust what you actually see.
seasalt
Having said that, I found Adword's reporting to be more accurate in the early days. At some point last year they made changes to this tool and it has been inaccurate ever since.