One more thing is that the offsets might differ between ranges, the offset in the 100mA range might be different from that of the 10mA range and hence it may not entirely cancel out.
You would have to use opposing currents for the delta method to cancel out these offsets as described in this article.
BTW, is there a reason you are using an SMU to measure a resistance?
I have done it before but most of our testers also have DMMs so unless I am doing something super specific (like having to use a pulse or having to do a high current measurement), I generally use the DMM.
Yes, it is both pulse and high current.
I found that the high lead resistance is the culprit for the inaccuracies. Without those, the 4141 and the 4139 work well. With the high lead wire resistance (4 Ohm), only the 4139 works well.
Thank you all for your help!