And just for grins and giggles I plotted out the trends for you. I did it with the CRUTEM3 (land) and HadCRUT3 (land and sea) temp anomaly observational data. Using both datasets I got an increasing trend from 1986-1997, from 1999-2006, and across the whole interval from 1986-2006. In fact, slope of the trend line after 1998 was more than from 1986-1997 (so a faster rate of increase). The only way I was able to get a decreasing trend line starting with 1998 was to stop at 2002 and even then the decrease was extremely small.
Here are the graphs for the HadCRUT3 data.
This one shows the trend line for 1986-2006. The red line is a 5 year running average.
This one shows the trendline from 1999-2006. With 1998 included the slope changes to 0.008 (still positive).
I also used the same data and plotted it out against atmospheric CO2 measurements from Mauna Loa for 1986-2003. You get an R-squared of 0.63. That means that changes in which of the two you assign as the independent variable (which would be CO2 based on the physics) could explain as much as 63% of the variability in dependent variable during that period. For 1999-2003 it's 0.81 (or 81%).