Moisture bias?

So after much discussion today about moisture there is accumulating evidence. Here is a comparison between the MWR on the roof (1.6cm at 00 UTC 5/17 and 2cm at 00 UTC 5/18) , the IPW from the GPS system (gpsmet.noaa.gov), and the radiosonde launch from Norman.
GPS:

Here is the sounding:

Note the PW is 0.5" with a characteristic decrease immediately off the surface. Just after this time the dew point at the NWC increased from 53 to 55 in the hour after release.

Below is the MWR PW time series with the last vertical dotted line at the right being 0000 UTC on 5/18 (the beginning of the chart is 5/17 at 00 UTC):

The MWR agrees well with the GPS system and the MWR is fresh off a calibration. I have no idea what would cause this discrepancy with the radiosonde data but clearly something is amiss.

Let me back up to DVN on the 16th at 0000 UTC for a counter example:

And now the corresponding GPS data from Rock Island:

Note the close correspondence for a few balloon launches; specifically at 5/16 00 UTC! Yet that decrease above the surface looks suspect. It is possible the cloud layer above 700 mb is laying a role but what I don't think we can diagnose.