Urban condo market analysis: noise or signal?

Here are two sets of urban condo data from highly reliable sources and that on the surface are difficult to reconcile:

1. The Warren Group "Boston" (defined as Beacon Hill, Back Bay, North End, and South End) condo median price for January over the last three years (available here):
2005 - $502,000
2006 - $364,000
2007 - $580,000
That works out to a yoy decline of 27.5% from 2005 to 2006, and a 59.3% increase from 2006 to 2007. Does anyone believe that the median condo in these neighborhoods increased in price 59% from last January to this January? An obvious problem here is the sample size. The total Jan. condo sales for each year were:
2005 - 172
2006 - 112
2007 - 143
If a new project or two comes online during a particular month, it can have an inordinate effect on the median price reported. To me these data are useless without more info about the projects that were coming online, and/or more data about the size and type of units that were selling.

2. Joe Wolvek of Coldwell Banker Residential Brokerage reports on his website (here) the price per square foot (obtained from MLS-PIN) for three of the four neighborhoods lumped together in the Warren Group "Boston" sample. Plotted below are the data for the past 3.5 years, plus the data for closed sales this year through 2/7:

Price per sq ft appears to have peaked in late 2005 / early 2006 for all three neighborhoods. A simple analysis of the quarterly YOY change in price per square foot results in the following plot:

The light blue line represents the MA statewide change in median condo price (not normalized to square footage). I added this line to show how closely these urban condo markets are tracking statewide trends when price per square foot is analyzed.

In contrast to the Warren Group numbers, there does not appear to be a significant surge upward in price per square foot in these neighborhoods. All three show price per square foot down 2-10% YOY so far in Q1 2007. Perhaps the North End is the critical piece of missing info that would reconcile the Warren and Wolvek numbers. Perhaps not. More likely the Warren numbers for nieghborhoods undergoing rapid development (at ever increasing price points) are subject to inevitable noise and sampling problems.