This is a very good point. Although the data (at this link) are not isolated, using a GUI window as a ruler, you can see a clear decrease, just not a visually compelling one. And yes, I agree that as you reduce the mortality associated with other diseases, getting cancer becomes more likely.
What would be more interesting to see, in this particular context, is the mortality associated with different disease groups for different age cohorts. For example, how has the treatment for cancer progressed for people between 40 and 50, rather than all cohorts at once.
The yellow band is also getting eaten from below by the blue (presumably AIDS deaths?).
US median age grew by about 2 years per decade between 1990-2000 so that also supports thisisnotmyname's point.
Some of the decline may be due to lifestyle changes (e.g. lung cancer has probably peaked in the US) but there has also been dramatic progress in treatment for some cancers (Breast, testicular, Hodkin's).
Also, some of the recent treatment modalities are genuinely revolutionary. For example, Gleevec (subsequently exposed problems aside) is pretty damn amazing. Obviously, new things are becoming possible because of a long-growing body of research. The question is whether we have reached the tipping point. As I said previously, I'm optimistic -- I think we have. That being said, cancer is fucking complex and even if we were at some tipping point, progress is measured in years and decades, not weeks as the breathless popsci articles always seem to imply.
What would be more interesting to see, in this particular context, is the mortality associated with different disease groups for different age cohorts. For example, how has the treatment for cancer progressed for people between 40 and 50, rather than all cohorts at once.