Gerhard,
A few additional comments:
1. I agree with you, not a good idea to use an evaluation board in TX mode to test another evaluation board in RX mode. It is always best to test a receiver with a test transmitter (RF signal generator) and to test a transmitter with a test receiver (modulation analyzer, or spectrum analyzer).
2. I notice the CC1125 RX sensitivity specs are listed as "typical" so perhaps TI will supply a worse case spec. It may be 1 or 2 dB worse.
3. For the 434 MHz RX sensitivity spec of -123 dBm (DEV=4kHz, BW=10 kHz): If I assume the NBW=10 kHz and S/N req=6 dB then NF=5 dB. A 5 dB receive noise figure is possible at 434 MHz band. Can you measure 6 dB S/N at the output of the CC1125 with some other equipment available to you and see what the sensitivity is?
3. As I wrote previously, power supply noise can sometimes be a problem for sensitivity. The lowest noise voltage regulator I know of is the TI LP5900.
4. If SW settings for the on board crystal or TCXO are in question, "hose" in a clock with a signal generator or TCXO on a breadboard that you can easily control as an experiment. If a signal generator is used then the phase noise will be poor compared to a TCXO but you should still be able to measure RX sensitivity.
Regards,
Eric Hooker
RF Consultant