There is some kind of a problem with our front end estimation of the cross power spectrum, we are still investigating. 

We found that there was a discrepancy in the inverse sensing gain applied to the sum and null channels for several days, including the time Craig is using above.  We also today found that the calibration of CAL_DELTAL external into meters saved in the dtt template we were using was out of date.  I've updated the calibration using the one that is in the DARM FOM, and saved the new cross corelation template in userapps/isc/h1/templates/ as well as updated the one in the noise budget directory. 

The attached plot shows the problem, the cross corelation seems to be below thermal noise at 200Hz, this was taken at a time when the settings in the front end were correct for the inverse sensing and there was no squeezing.  The blue trace is the DARM noise according to GDS calib strain, while the purple trace is based on CAL DELTAL with the transfer function in the DARM FOM template applied, there is a discrepancy around 200 Hz between the two traces.  We will try to investigate further by running the LLO cross corelation script.