As you can see the amplitude starts dropping after 10MHz and at 20MHz it is 720Vpp. Is this normal? Could someone reproduce it?
You are not crazy.
EDIT:
This isn't some aliasing issue with the scope. When the timebase is set to something more common, say 100nS, it is possible to
see the sine wave shrink in the vertical dimension (voltage) as the frequency approaches 20MHz (when the DG1022 is set to sweep). In static frequency mode, changing the frequency on the DG1022 (using the rotary encoder) from 1 Mhz to 20 MHz in 1 MHz steps results in the output remaining between 1.01 Vpp - 1.03 Vpp.
TEST SETUP:
Rigol DG1022
Firmware version: 00.02.00.06.00.02.07
CH1 output set to 50 Ohm
Sine wave, 1Vpp
Sweep:
Start: 1Hz
Stop: 20MHz
Time: 3s
Rigol DG1022, channel 1 --> Amphenol 50 Ohm BNC-BNC cable, 36" (~1 meter) --> Amphenol 50 Ohm BNC-BNC terminator --> Rigol DS1202Z-E