Well...
What's the setup? BNC to BNC? 50 ohms source and load? Or no termination?
What's the generator bandwidth? It's in the datasheet.
What's the scope datasheet? It's in the bandwidth.
(Did I really just type that? Ugh..
)
If using a probe, what's the bandwidth of that, and under what setting (if it's 1X or 10X or switchable)?
Can you come up with a test that might prove the bandwidth of one instrument or the other, independently?
Example: an RF probe with diode, capacitor and resistor. Without calibration, and with reasonable construction, you can reasonably assume it will behave fairly well at most any frequency the diode is happy with.
Can you come up with reasons that such a back-up test might fail? For instance, a spiky distorted waveform will read higher than it should.
And ways to test that, in turn? For instance, testing the generator at just one frequency, with an LC resonant filter to remove nearly all distortion, and using that to provide a higher confidence data point with the RF probe. Or using the filter to isolate harmonic distortion, and verify that the waveform is clean like you want (generator at 10MHz, filter at 20MHz (and so on): little or no voltage output = a good thing).
Tim