Nope, you're missing nothing. That's absolutely correct. The problem (from merpnderp's point of view) is that the data for the past ~20 years doesn't differ significantly from the 30 year (or longer) trend. It's using the lack of significance to mean something significant which is the root of the error.