This is email from Dennis Keyser at NCEP this morning:

All versions of the PREPDATA data cards were modified at 12Z on 07
January in order to flag moisture data from the Gulf Stream dropwinsonde
flights. This means that specific humidity on all levels will be
flagged with quality mark 15 in the PREPBUFR files, and the moisture
data will not be used by any NCEP analyses (including the RUC). (In the
OI-version, the specific humidity on all levels is set to missing for
these data.)

James Franklin of NOAA/AOML/Hurricane Research Division has indicated
that these flights are again showing a dry bias in RH. There will be
flights beginning the second week of January.