When you put a 1000 ohm resistor between CH0 (port1) and CH1 (port 2) of the NanoVNA you should measure an "Insertion Loss" of -20.8 dB if the NanoVNA was a perfect device. It is not for several reasons including impedance of CH0 and CH1 are not exactly 50 ohms and the way calibration is done. So your measurement of 20 db is pretty close.
Your LTSpice and Sig Gen experiment is off by 6 dB because you did not calculate correctly. Try calculating the power in dBm of the 50 ohm load resistor without the 1K resistor and with the 1K resistor. Subtract the two and you get a 20.8 dB difference.
Roger
Roger Need