For faint objects, use the high gain setting (fewer electrons per digital number). This will sample the readout noise better, although it limits your dynamic range somewhat. Low gain is sometimes used for very bright objects. However, you should probably stick to the same gain during your run.
For HIRES, the electrons per digital number (eperdn) is set according to ./makee/eperdnHIRES_higain.dat (eperdn=2.45) or ./makee/eperdnESI_logain.dat (eperdn=4.80). The readout noise is set according to ./makee/ronoiseHIRES.dat (ronoise=6.0 electrons). You can override these values by editting these files, or giving "eperdn=" or "ronoise=" on the MAKEE command line.
The alignment of the cross disperser angle (XDANGL) between the flat image and the object image is very important. makee determines the slit length using the flat field orders, so any misalignment can severely decrease the effective length of your slit. This can be especially bad if you have overlapping orders in the blue. A change of 0.001 XDANGL units is about 1 row (unbinned). The makee XDANGL default tolerance is about 0.002 units, although in practice the program can be run with larger differences in XDANGL.
It is especially important to take an arclamp exposure next to the object exposure when using the UV cross disperser, since there are no usuable night sky lines to obtain an accurate wavelength scale shift. In later releases of makee (1999 or later), the program can "average" the wavelength scales of two arc lamp exposures taken immediately before and after the object exposure. This is the preferred strategy when using the UV cross disperser.
A reasonable ThAr exposure time is 1 second with the clear lamp filter. If too many lines are saturating (more than 4 or 5 in any order), you may consider using the ng3 lamp filter. In the red, beyond about 7000 Angstroms, there are many very bright Argon lines which will saturate and bleed down columns. This is normal and the makee programs are designed to ignore these lines.
Only the column-to-column bias pattern is used (row-to-row variations are ignored), so makee will obtain the bias pattern by median-ing each column. This takes care of any cosmic rays, so it is only necessary to take one good short dark exposure.
You can also use a pinhole quartz exposure to define the trace function. This avoids having to take a star exposure and should work just as well as a star. Just use the same setup as your flat field exposures and then put in the "D5" decker. You can also probably go 2 or 3 times longer in exposure time to insure you have enough counts in the blue (at least several hundred at any column position across the entire order.)
Here is a list of "good" HIRES standard stars: standard star list . A 300 second exposure is usually adequate for these stars. For more on flux calibration, see: flux calibration.