I should also have mentioned that the FGs own counter is also off by 5%. So, when I feed the 1k signal back to the FG it reads 1k. If I feed an external 1k signal (calibrated by my DSOs hardware counter) the FG reports that it's off by 5%. So, this issue affects the signal generation *and* the counter by the looks of it.
They're both based on the same internal clock, so they'll both be affected.
I haven't heard of one of these being that far off, not that I've seen a large number of these generators. However, if you have access to a GPSDO, Rubidium, or other high-stability oscillator or a counter with one as its external clock, that'd be a better basis for verifying your signal generator's accuracy.
If you're using an oscillator directly, use your scope to compare its output signal to that of the 3224 at the same frequency. You'll see how much drift there is between them. Alternatively, if your scope can accept the oscillator's signal as an external clock, then that'll improve the scope's accuracy for measuring the generator.
With a universal/frequency counter that has a highly stable external oscillator, you can directly measure the generator's output frequency.