Given the cost of these modules I would not expect miracles, and would say just 1dB loss is great.
So 1db lost in cable - you know about that, hence you expect ~19dB.
Assuming analyzer and procedure are perfect, I fully expect another 1-1.5dB is lost in translation (ie. RF switch, impedance matching imperfections, PCB signal routing discontinuities). Think about how many twists and turns that signal has to go through from exiting the TX pin of the transceiver chip, and into the port of the analyzer. So you loose 1-2dB which lands you at 17-18dB you first mentioned you get.
But since you're chasing that last dB - do you actually need to reach full 20dB?
As soon as you add the elephant variable - ANTENNA! - that starts attenuating in ways you cannot fully control (mostly based on antenna type/parameters).