Have you read the details on the picture name?
OOPS! Totally missed the picture name.
Then I stand corrected. It's a 50 mV variation from 1 to 30 MHz on a ~2.5 Vp signal.
I.e, 20 x (log10(2.525 - 0.05) - log10(2.525) ~= -0.174 dB , a bit less than my estimation.
Kudos to the output stage designers in Siglent (or FW designers for the calibration table, or both).
But I think this is a good lesson for better probing, connections, and termination techniques.
Don't worry, we all were wet behind the ears, once: just three days ago a dear and usually really smart friend of mine was probing a small RF power amplifier prototype.
He sent me some screen capture - I could not understand what was happening - for me it had worked with no issue...
Then he casually mentioned that he was not using the probe spring.
"Oh, definitely do not use the alligator for RF or fast edged signals!"
"No worries, I'm not using that either!"