It synchronizes quickly and all readings are ok except Power.
I get out of tolerance. I can effect this by changing PCL.
My question is how to understand the plot? The plot shows power in the time domain
but it looks like the unit is out of tolerance at the start (pre-zero) and end of the sweep.
There are obviously spurious emissions above the specification detected. Either a real problem of your GSM device, or just an unsuitable measurement setup, e.g. receiving environmental RF due to insufficient shielding.
From the picture you posted, it looks that your burst is touching/overlap the (red) specification mask (in few places I see).
There are multiple reasons for doing this, but first that I would check is the path loss settings of the CMU200. Verify the settings for the attenuation loss between CMU200 RF connector and your GSM device antenna connector.