I just realized that the data from the CN1 project contains the answer to your question.
From the data it can be seen that the attenuation ratio of 6.63E-2 (1x-range-2) works just fine, so all other ranges except 1x-range-3 should be good to go with respect to attenuation, because they have less attenuation before the ADC. My unknown is the resolution of the ADC and that would affect the largest useful V/Div in each range. But then again who knows what size of signal is on that DC voltage, maybe it is much larger than mV, so maybe there is not a known upper V/Div limit to consider. The user through trial and error will find that limit for his particular configuration.
[code]Ra Rb Req atten ratio
5.10E+05 5.10E+05 5.00E-01
5.10E+05 3.90E+04 3.62E+04 6.63E-02
5.10E+05 3.30E+03 3.28E+03 6.39E-03
5.10E+05 5.10E+05 5.00E-01
5.10E+05 7.50E+04 6.54E+04 1.14E-01
5.10E+05 6.00E+03 5.93E+03 1.15E-02
[/code]