The real short % according to S3's data is 122%. However, their 55% figure is technically not a lie, but extremely misleading. I will explain everything.
Here is what they did:
Sources (S3 head):
S3 head is redefining share float to include shares that don't exist in order to be able to say shorted % of float is lower.
it reduces the traditional SI % Float, Instead of Shares Shorted/Float our calc is Shares Shorted/ (Float + Shares Shorted)
So, by this definition, if a stock is shorted 400% of existing shares (total banana count borrowed and resold 4x) and total shares is 100, short % is calculated like this:
400 shorts / (100 shares + 400 longs whose shares are borrowed) = 0.8
That is, the normal way we define short % would say it's 400% shorted. S3's way says 80%.
Knowing this formula, we can work back to what S3 would have said the short % of float was using the normal definition of short % of float:
55% short of float means for all existing shares + shorts (or, ont he other side of the trade "longs whose shares were borrowed away to short") is 55/45 as much as existing shares. Meaning, portion of shares short by the normal definition (% of existing bananas borrowed) is 55/45 = 1.22
That is, S3's data is telling them that after friday trading, GME is still 122% short.