R&S CMU200 problem with connection

Hello,
I work with CMU200 BTS simulator. I have a problem: I can't connect with phone when attenuation of coupler antenna is more than 25dBm. For example: I work with GSM900, TX power set to 20 dBm, RX power to -50dBm, phone lies on the coupler antenna, connecion is established, CMU gives me informations: TX 12 dBm, RX: -65 dBm. So there is some attenuation. Whene I move phone away from antenna (f.e. 20cm away) then of course TX and RX have worse values, for example TX -25 dBm, RX -85 dBm, but connection is in progress altthough TX losses of 45dBm.
But if phone is not connected and I move it few cm, then connection is not possible to establish. It is always, when TX losses are more than 25dBm. So if TX settings is f.e. 20dBm and CMU200 show -5, -5,5 and less then connection can't be establish. But phone see BTS (CMU200), because it has almost full range(coverage). So phone see BTS, but BTS doesn't see phone, this two devices can't synchronize. With other BTS simulator there is no problem, connection is established altthough losses of 40 50 dBm.
Where can I find some settings for this limit? Why I can't establish connection whene attenuation between BTS and phone is more than 25dBm?

P.S.
Sorry for my english

For Radiated Test using CMU200 always use RF2 connector of the device. RF2 input/output has higher dynamic range than RF1, and for radiated test you need higher dynamic range.
Try to increase the level of the BCCH and of the TCH to about -50dBm each, and set the path loss on RF2 (in Connection Control) to about 20dB initially. You can increase (or decrease) this loss later, during the measurement.
Beware that R&S instruments use positive numbers for cable (or path loss), when for example Agilent test devices for use negative numbers for setting external losses.

Check this R&S document for GSM measurements using CMU200:
**broken link removed**

I solved this problem myself, I add some gain (virtual gain? I'm not sure It was physically added or only added to calculation, f.e. 10 or 20 dBm) to RF2 port (for TX only), and when I was writing measurements results from CMU200 I subtract this gain (10 or 20 dBm), but everytime connection was established, because attenuation wasn't more than 25dBm.