As discussed in the FAQ: What is the data output coding format used in AD717x family?, the AD717x uses unipolar coding with input range from 0V to positive Full Scale and bipolar coding with input range from negative Full Scale to positive Full Scale.
It is recommended to use bipolar coding regardless of whether the input is unipolar or bipolar. In unipolar coding, if the actual input is a negative value for example due to offset error, then that ADC will treat it as 0 code because the ADC is not able to convert the negative input voltage, so you are now blinded with the actual differential input and mistaken it as a perfect 0V. With bipolar coding, the ADC can convert negative differential voltage instead of just a perfect 0V when using unipolar coding.
One thing to note is that regardless of coding format used, same effective or peak to peak resolution will be achieved. In AD717x, since the ADC noise value is larger than the LSB size, so the effective resolution or peak to peak resolution will be limited by the ADC noise. You may refer to the equation below for effective and peak to peak resolution.
