Thank you, that link (particularly [1]) is basically the kind of explanation I was looking for in my initial comment.

[1] https://physics.stackexchange.com/a/784886

There's so many variables there! What kind of purity could 1889 achieve? What kind of uniformity between the N samples? How do they know the different cleaning procedures (!) are not adding mass by leaving something behind? If the samples are known to be oxidizing, why leave them in air? Why are there multiple elbows in that graph around 1950 -- surely another procedural change but again not uniformly applied?

Maybe these aren't the best sources of historic mass data.