control accelerometer averaging?

Hello, I am new to this board, but not to vibration theory.

Can someone please shed some light on what would be "allowable" limits on individual control accelerometers? Say I'm running a vibration test with an average of 4 controls. What, if any, would be the limit on any individual accelerometer? For example, if I were running a 15 grms test, is it acceptable to have one control reading very low e.g. 2 g's and another very high eg. 32 grms, etc., as long as the control average looks good and within tolerance? What limits would MIL-STD-810F allow? Can one accelerometer be essentially "zero" with another @ ~90 g's as long as the average looks OK?

Thanks in advance for any inputs or comments you may have!