The value of 0.2 nm is a sampling uncertainty. Call it S. What you also need is the device precision (instrument resolution). Suppose, for a given set of conditions, the value is X (nm). The total uncertainty in any measurement made under exactly the same conditions is U^2 = S^2 + X^2.

You have two average value V1 and V2 where the only variable is sample. You want to make a statistically rigorous comparison of V1 and V2. To do so requires accurate values for U1 and U2. The value of S may be different for each sample. This is the randomness in preparing or measuring the same sample for example. What you hope is that X remains the same independent of sample. You hope that X is only dependent on the instrument, and, since you use the same instrument and instrument settings, X is the same even when you change the sample. You allude to a problem in carrying forward on this assumption by your statement about tapping mode versus contact and ScanAsyst modes.

I can agree that, in tapping mode, X may depend on the type of sample. This can be as simple as having a hydrophobic versus a hydrophilic surface. Measurements of heights can be less precise on one versus the other in tapping mode. In contrast, the value of X for height measurements in contact mode can be taken to be (nearly) invariant from sample to sample. I say nearly only because the assumption is limited to scaling the geometric size while keeping form factor exactly the same. You could have issues measuring heights over vertical steps versus over curved steps, for example.

Thanks a lot for the detailed explanation. We have two different AFM's: one with height noise X

It is interesting and illuminating simply to calculate how many atoms of a given element or molecules of a given compound line up to make 0.2 nm....

Actually, we are trying to interpret some structural changes in the molecules causing this difference. The small height difference with our molecules of interest is usually not due to no. of atoms, but their orientation. Theoretically, two atoms (C or N or O) are sufficient to cause a height difference of 0.3 nm.