My suggestion from another thread several years ago:
During the Rutherford years, the memorial totals were representative for the US (where companies tended to be quite large) but only counted figures with more than 20 or 30 people at each location. JWs were still rather thinly distributed in foreign countries; the 1935 Yearbook shows that in the 1934 international campaign, there were 41,000 publishers in 83 countries (p. 44) -- an average of about 490 JWs per country, and within each country the memorial gatherings were probably local. The original 1935 figures clearly underestimate the true number of JWs in other countries, as is evident from the subtotals on p. 53 in the 1936 Yearbook and the 10,537 attendees mentioned on p. 120 for Central Europe (i.e. Austria, Belgium, Danzig, France, Italy, Luxembourg, Poland, Romania, Saar Basin, Switzerland, The Netherlands, and Yugoslavia) which exceed the subtotal for international memorial attendees on p. 53. Indeed, the 1935 Yearbook (p. 99) indicated that there were 450 companies in Central Europe, so there were only an average of 23 memorial attendees per company. If the international totals counted only companies that had a minimum of 30 attendees, it is reasonable to expect that a majority of the attendees were not counted.
The revised figures of 1935 attendees and partakers first appeared in the Proclaimers book (p. 717), so I suspect that they were revised as part of the research that went into that book. On this view, the writers undertook the painstaking work of retabulating the attendance figures from records of all the memorial gatherings around the world in that year. So the revised numbers are probably more representative of the true number of memorial attendees, and are thus more comparable to current memorial figures which are counted the same way. But the revised numbers are not comparable to the figures from the 1920s which were tabulated in the old-fashioned way. If you are interested in comparing the 1930s attendance totals with the 1920s attendance totals to gain an insight on changes in membership, you would probably need to use the original figures which were tabulated the same way, even though they problematically underestimate the true number. If you examine the 1937 Yearbook (p. 69), you will find similar international totals (35,172 attending and 25,435 partaking) to the one in the 1936 Yearbook. Interestingly, the next several Yearbooks declined to publish any international attendance totals, although the 1958 Yearbook reported a total of 69,345 for 1938 (p. 284) -- a number that is quite comparable to the 63,146 in the revised international total for 1935 in the Proclaimers book. This suggests that it was in the late 1930s when the Society adopted the current, more accurate way of reporting memorial attendance totals.