Maybe I should add that maximum external voltage will be 5.10V. That is, after reaching 4.85V, the source will keep rising until it reaches 5.10V tops. I won't raise anymore after that. So a resistor divider of the magnitude you've proposed is too much of a stretch.If you divide 20 volts down to 1 volt using 190K and 10K resistors then that meets your 10k source resistance requirement. Only 100uA is drawn from your 20 volt source.
Placing a cap from the A/D input to ground also helps at higher value source resistances.
Let me add some numbers here.... if I want to be able to measure 4.85V with a margin of error of ±0.01V, then I'd need more than just an 8-bit resolution ADC ... on the other hand, I could perfectly live with an arrangement that could detect anything between 4.85V and 5.00V ... so that's a 0.15V working margin for you ... maybe there's a simpler way of accomplishing what I want under that last consideration ... but I can't see which ...