Ok thats all good information to get in ,
Word from the mouth of someone with 60 years experience in electronics is you use finer tollerance resistors to make up the potential divider across the output IC . Each IC has a range of specific resistor values ,not simply any resistors of a certain ratio,that ensure best performance ,generally between gains of about 2 and 5 best performance is had .
When the drive to the 3095 is engaged the signal is fed from the opa686 outputs , my guess is theres an offset built into the software calibration which tries to make up for the slight gain differences as you switch between high and low range ,if we have calibration 'flat' and can adjust amplitudes independantly between opa and ths op amps externally we could essentially do away with using some calibration features in the digital domain ,maybe this is the issue as when you switched to 5.1 volts the output level actually went down DPA31.
So were essentially getting 12 bit performance , good op amp technique should be able to squeeze those last two bits out easy ,better sorting the amplitude differences between ranges in the analog domain as much as possible and not relying on bit munching internal calibration .
Theres a few nice theories for you to mull over in any case ,
Your idea to piggy back two resistors has merrit too Dom , if you picked x2 the required values ,then selected from a bunch of 0.1% tollerance in both rg and rf , you could get the resolution of the display to read bang-on to better than what we see now , if the range was a few mV off you could simply select from your bag of components a marginally lower value for each channel individually until its right 'on the money' ,in the words of Zappa .
Kindest regards Gents ,
Its Friday and the beer is flowing ,I hope your taking a dose of your fav poison and kicking back like me ........
as is as we saw its maybe a few tens of mV off across the ranges at a volt or two amplitude .