How close compensation is good enough?

Dear friends:
I have been trying to tackle two compensation issues, and I am asking for 
your help to get them straight.

I am working with double (or triple) stains in order to determine 
correlations between marker expression levels, for example, to see wether 
the HLA-DR high cells express CD86 more than DR low cells. I need to set the 
compensation as close as possible, so my question is how close is good 
enough? I use single stained cells as the comp controls so they have the 
same autofluorescence and I use the brightest stain to set the compensation. 
If for example the median in the FL2 channel of unstained cells is 4, what 
would be the range of acceptably good compensation of FITC stains?

The other question I have is how much does imperfect (under or over) 
compensation affect the fluorescence values obtained? I think the effect 
will be different for different FITC stains with different intensities and 
thus different amounts of spillover, but I have a bit of trouble translating 
that idea into numbers. for example, if the FL2 channel is undercompensated 
by 5 median units (FL2-FL1 % too low), how much will that affect the median 
in the FL2 channel of the double stain FITC-PE? will it "increase" it by 5 
units as well or is the relationship more complex? is there any way to 
calculate or estimate the range of the effect?

Thanks in advance for your help,

Sincerely,
Uriel.
MD/PhD student
The Laboratory for Cellular and Molecular Immunology
The Hebrew University - Hadassah Medical Organization
Jerusalem - ISRAEL