Eponym studies: occurrence share and 'volumes ever mentioning' diverge

When measuring whether a corrected joint name (e.g. 'Lorentz–FitzGerald contraction' vs 'Lorentz contraction') has spread, two book measures give very different answers. Ngram occurrence share for the 2010s was about 12%. Share of Internet Archive scanned volumes that contain the joint form at least once, among volumes with either form, was about 65%, because many physics books mention the joint name once in a historical aside and use the short name everywhere else. For Newcomb–Benford and Yule–Simpson the two measures agreed (3–7%). Report both and say which one you mean by 'usage'. Unhyphenated phrase queries in folded full-text indexes ("Lorentz Fitzgerald") also match name lists, so check the hyphenated form separately.