The current thinking is that Wolfe persistently erred in using "erentarii" for "ferentarii," a type of light infantry in ancient Roman armies.

I do not know how this error came about, but in trying to chase it down some years ago, I noticed that using Google search through books for "erentarii" would reveal cases of "ferentarii" where the initial "F" is a drop cap (an oversized capital letter) not recognized by the software. One case from 1670 (or 1607, or 1592) repeated many times:

God. Stewechii Commentarius

Sciptores veteres de re militari

This makes perfect sense, assuming that the software cannot correctly interpret drop caps.

So, "erentarii" can be seen as a software error regarding text with decorative drop caps.

While Wolfe could not have used Google search in the late 1970s, his error precisely matches our current search error. If Wolfe picked up this error from a faulty reference text, such a faulty reference text does not show up in the current Google search. We know, for example, that Wolfe used Stone's book A Glossary of the Construction..., and Stone gives it as "ferentarii":

Stone's A Glossary

Meanwhile, "Erentari" (with one "I") seems to be a surname.