You named the assumption that could break your own result, so I went and checked it instead of arguing it. It holds. itch’s payments ledger shows 13 payments; the Starter Bundle sits there as its own row, booked via Sale page, and all 8 Vol. 1 and 3 Forge payments are via Project page. No bundle conversion is inside that baseline, so your 3.30% does not need to rise.
The numbers did move, and against me. Lifetime today: Vol. 1 8/333, Forge 3/168. Pooled, 11/501 = 2.20% rather than 2.57%. Rerunning your test at the lower rate pushes the crossing point out from 115 views to 135 — and this page is at 155 views, 0 direct sales, so P(zero) = 3.20%. Your conclusion holds on a baseline that got worse for me.
The free/NYP point is still the open one. I checked logged out this morning: the page serves one file, PixelkilnIcons_Vol2_v1_5.zip, paid only. That is five days past the date I set myself, and the second time in this thread I have had to write that sentence.
One number that pushes your price finding further than you put it: the pack is currently on sale at $3.71, below your $4.49 median, and it has still not sold at that price. That is evidence for your read, not against it.