Crisis of conscience- sort of

Greetings all-

I have a situation on my hands that is driving me nuts. My therapist
tells
me not to let it bother me, but it does.

Here goes:

I have a structure from a 10 x 10 x 120 micron U-Oxide material. Small
crystal, but long counts on an APEX did the trick.  The space group is
Pmcn,
and the cell is 7.14  15.52  22.31.  My experience with these materials
is that
absorption is a headache no matter how you treat it. Thus I usually go
with
the most reasonable and painless route- this time I chose  SADABS.  This

gives me a best Rint of 10.4  and R1 = 5.7. Not bad for these types of
compounds. One oxygen atom in this model cannot be refined
anisotropically.
 This isn't such a big deal- everyone working in U-O materials has had
to
use the ISOR command at some point. However, in an effort to clean
things
up a bit, I cut the data off at 0.8 (Mo radiation).  This gives me an
Rint
of 9.5, R1 of 5.1 and TWO non-positive definites.  I then tried to
re-address the absorption by processing with SORTAV inside WINGX and the

empirical absorption correction therein.  This looks a lot like SADABS
with
a few more things to tweak.  The best I could come up with was Rint =
11.0,
R1 = 6.55 and NO non-positive definites (including partially occupied
H2O
sites!).

So, my dilemma is thus- which data treatment do I go with?

Option 1: decent R1 (5.7%), acceptable Rint (10.4%) and one NPD
Option 2: cut off data and get a decent R1 (5.1%), better Rint (9.5%)and

TWO NPDs
Option 3: acceptable R1 (6.5%) and worst Rint (11%) and NO NPDs.


By the way- I am very confident with the space group choice as I have
put
this structure through the ringer searching for pseudo and missing
symmetry.

Comments, anyone?
Thanks
Chris Cahill