It comes down to what qualifies as an attestation. If someone with "street cred" writes an essay for the Sunday paper and says, "I can guarantee you we used [ some word ] when I was a kid in the late 1940s early 1950s to mean such-and-such, and then letters to the editor arrive corroborating that claim, a dictionary might date the origin as "circa 1950" even if the first written attestation did not appear until, say, 1966. It's an editorial decision. For words from the middle ages, how would anyone know?
That said, when there's evidence of, say, a bunch of similarly structured verbs attested uninterrupted from the 9th century through to the 15th century, all going through similar sound changes, but there's one verb of the same ilk that is attested in the 13th, 14th, and 15th centuries but unattested in early medieval texts, an etymologist might well posit the unattested form, putting an asterisk in front of it as a sign that it is unattested, but indicating that there's hardly any doubt it did exist.
The historical record of the language is imperfect. Many manuscripts were destroyed, intentionally or unintentionally.