Factored loads are inflated loads. That is, each type of load has a specific safety factor (load factor) added. If you calculate settlements with inflated loads, the calculated settlements also will be inflated incorrectly.
I think the question is; how are you going to address uncertainty, or in other words safety factors, with respect to allowable settlement.
Why not use both approaches?
Unfactored soil properties and unfactored loads is your best estimate of actual settlement. This is informative, but is unlikely to be the deciding criteria. Settlement could be a bit more if your load estimate is a little out or if the soil properties are a little worse.
Factored load and factored (or lower bound) soil properties are a check to see that everything is acceptable in the same way you would use a safety factor on bearing capacity. If everything is done correctly "real settlement" will be less then this figure and closer to the unfactored estimate.
The remaining question is what factors to use. This will require consideration of the local design codes and other factors best dealt with by an appropriate geotechnical specialist.
The rationale for using factored loads and cautious estimates of soil properties is to achieve a desired level of reliability.
In the European codes (Eurocodes), both factored loads and cautious estimates of geotechnical values are applied (characteristic values). Factors are different though when dealing with limit states and serviceability states.
I agree that sometimes the results are overly conservative, against evidence of existing buildings showing no signs of significant settlements. This happens, I noticed, particularly when dealing with the edometric or other compressibility moduli and long-term settlements and may be possibly due to procedure(investigation and lab) biases or overconservative literature correlations.
