The easiest way to imagine it is to place a 3dB attenuator on the CH0 port and measure the Return Loss with nothing connected to the other side of the attenuator so that is open circuit.
The power will flow one way and be subject to 3dB attenuation (approx) and then reflected back from the open circuit and experience another 3dB of attenuation, making it 6dB in total.
The same happens if the other side of the attenuator is short circuit.
With higher values of attenuation, the actual value, when incorrectly terminated at one port gets closer to the theoretical value, so there is a closer relationship between the measured RL as being twice the value of attenuation.
Regards,
Martin - G8JNJ
