Resolution Index, Settling Factor, Noise and Effective Resolution

I am using LabJack U6 to prototype a design, which I will then need to implement on my own hardware. I am using LJStreamUD to gather data.

1. Why does LJStreamUD data have more noise than LJLogUD data. I know quoted LJLogUD noise assumes the 24bit ADC, but if it is only a difference between which ADC one uses (16bit or 24bit) then why does the 24bit ADC have less noise. The 24bit ADC (from clause 3.1 of the data sheet) is quoted as having 16 effective bits when the resolutoin index is 1. The 16bit ADC (from clause 3.2 of the data sheet) is quoted as having plus/minus 3, 16bit counts of noise, which translates to 13.4 effective bits.

2. What is changing the resolution index doing internaly. My guess is that it is increasing the oversampling rate and that resulting samples are being averaged, but there seems to be no standard relationship between the resolution index number and the amount of oversampling that might be taking place. If it is oversampling that is taking place, what is it for each resolution index?

3. I presume Settling Factor is affecting the sample time somehow (before the conversion actualy starts). How specificaly does Settling Factor affect sample time? What is the size of the sample capacitor? What is the internal resistance between the sample inputs and the sample capacitor when sampling?

Thank you.