The technique of subsetting the data according to housekeeping parameters is in its infancy. The advice for now is :
Background reduction
--------------------
To minimise the number of particles in the background -
select EE_MV (for MPE PSPC) to be between 0 and 170, this tends
to lose 
10% of the data.
NB: This parameter is only recorded for PSPC data. No information is currently available on minimising the HRI background.
Aspect improvement ------------------ Some on-axis sources exhibit strange PSFs, with a characteristic lump in the bottom right quadrant. This can be improved by setting constraints on the aspect error - set ASP_ERR to be between 0 and 2 (the units of ASP_ERR are arcsecs/2). This parameter may be used for both PSPC and HRI data.
NB: Data received in US FITS format isn't quite as useful. The aspect error varies between 0 and 5, with everything over 1 being a large error (I think, info. is hard to come by !). They also use -33 (and/or 9) to mean missing aspect info.