Strange GMSK behavior when transmitting it on 868Mhz at 250kbaud

I am trying to send packets from a USRP to a CC1101 dev board. I was having trouble getting it to work, so I hooked up my USRP to a spectrum analyzer. I'm transmitting GMSK on 868 MHz at 250 kbaud. Here's a video showing the spectrum analyzer capture of that:
http://www.youtube.com/watch?v=dmjkx...ature=youtu.be

I noticed that the constellation diagram is jumping around, and the spectrum seems to get saturated. I'm transmitting continuous 0xA's from the USRP, but that's not what's being demodulated on the spectrum analyzer there. Does anyone see anything really wrong with that capture? I'm still learning this stuff, so it's possible there's an obvious reason I'm having problems but I'm missing it.

Looks like a dynamic DC offset problem. This could be from multiple causes, including layout.

Do you think that could be at least partially responsible for my problem? And what was it in those plots that told you there was a DC offset?

Your video capture is not very long, but I see in the spectrum plot, that the floor goes up about 40dB's when the IQ circle plot goes off-center.