To be fair, we should be seeing the same kind of drop percentage wise in the LuckyBancho surveys. We don't. It shows a decline yes. But once you factor out changes made to privacy settings on the Lodestone.
To preface this; I'll be using two iterations of the Dawntrail numbers for this from LuckyBancho's surveys. I've translated the images for this thread for reader convenience. I've linked the originals as well.
These numbers cannot by definition go down, as you cannot "unclear" Dawntrail or "unown" the Dawntrail Early Access bonus.
LuckyBancho Numbers as of August 27th, 2024 (https://luckybancho.ldblog.jp/archives/58615522.html) :
LuckyBancho Numbers as of December 29th, 2024 (https://luckybancho.ldblog.jp/archives/58883226.html):
The data isn't 1:1 comparable however, due to changes to how the Lodestone changed privacy settings in 7.0 (according to LuckyBancho). Thankfully, we can adjust for this. We see a drop in players who had Dawntrail Early Access or had cleared Dawntrail between the surveys taken on August 27th 2024 and December 29th 2024.
On August 27th, 2024,
1,092,592 people had Dawntrail Early Access. On December 29th, only 785,035 people had Dawntrail Early Access. I doubt Square Enix gave out 300k refunds of Early Access between 2 months after expansion release and 6 months after expansion release. It's pretty safe to assume these were changes in privacy settings. I would assume this would mean a change in a default for these settings, though, as I also find it hard to believe that 29% of the playerbase changed their privacy settings.
That being said, with the recent things about the account ID third party tool could explain why more people than you'd expect changed their privacy settings.
In total, we see a 28.16% drop in Early Access numbers being counted. As Early Access was long over by August, this number should be constant.
Percentage Drop = [(1,092,592 - 785,035) / 1,092,592] * 100
Percentage Drop ≈ (307,557 / 1,092,592) * 100
Percentage Drop ≈ 28.16%
Ratio = 1,092,592 / 785,035 ≈ 1.39177488902 - This is the factor that we'd need to multiply the December numbers with to get somewhat comparable data.
And sure enough, if we multiply 1,092,592 with 1.39177488902, we get 1,092,592.
So, multiplying the Active player number from December 29th with the factor gives us
1,003,091 * 1.39177488902 = 1,396,076.87 ≈ 1,396,077.
And let's compare this to the August number: 1,440,997.
That is indeed a slight drop. But it's only by 45k players. That's a 3.12% drop. Not a 75% drop like the OP would suggest.