IP3 of SA44

I have been experimenting with a couple of RF generators and my SA44, in an to benchmark the unit for doing some IMD testing on some RX and TX. I found I was able to achieve close +35 dBm IP3.(Third order Intercept Point)
I added some extra attenuation between the SA44 and the combiner (homemade), but no difference, so I have some more room to play.
My suspicion is that the Signal generators are talking to each other (via their ALC circuits)despite having close to 50 dB isolation.
The tests were done at close frequencies of a few kHz separation, and when I separated the tones by about a MHz, the IP3 worsened considerably, possibly due to different processing by the SA44