[Mark McLeod]

That cross-year comparison of % might be deceptive and I'm thinking about other ways to portray this kind of data. Maybe normalize % against the total number of teams each year, so losing 1% of total teams in 1999 can reasonably be compared to 1% of teams in 2012.
For instance, 1999 lost 12.7% of it's teams, while 2000 lost almost 5% fewer (7.9%) of it's teams.
P.S. I added full comparison charts that show the difference.

The "problem" with 2000 is that it had the lowest overall dropout rate of all the years charted, so the percentages are disproportional in comparison to the surrounding years. 2000 only lost 32 teams.
Most of the other years are pretty similar with less than a 1% spread in losses of total teams (7.9% - 8.8%), so the charts do generally work in comparison. The outliers, where comparison doesn't work, are 1999(12.7%), 2002(10%), and 2005(11.9%), all poor years for retaining teams.