David Calabro (personal communication) wonders if a conjunction and might be missing from the compound number here—namely, he suggests that the original text might have read “about three thousand and five hundred souls”. The loss of and could have resulted from the scribe’s eye skipping over the and because thousand itself ends in and. Of course, if the and had been written as an ampersand (Oliver Cowdery’s practice), that single character could have also been skipped as the scribe copied from 𝓞 into 𝓟.
Elsewhere the Book of Mormon text has one (and only one) and in every compound number involving thousand:
The example in 3 Nephi 17:25 is very similar to Alma 4:5, yet in that instance there is an and after the thousand.
It is important to keep in mind that in English the and is not necessary; in fact, in modern English it is not expected. Of course, one could argue that the expected lack of the and in modern English may have been the reason the and was lost here in Alma 4:5. But since the and is not required, the critical text will maintain the earliest reading without any and here in Alma 4:5 (that is, “about three thousand five hundred souls”). This is a unique reading in the text, but unique readings do occur. Of course, the possibility remains that this reading could be an instance where and was accidentally lost.
For another example where and may have been lost from a compound number, see under Mormon 2:2. There the printer’s manuscript reads “three hundred & twenty six years” while the 1830 edition reads “three hundred and twenty and six years”. For this part of the text, both 𝓟 and the 1830 edition are firsthand copies of 𝓞.
Summary: Accept in Alma 4:5 the reading without and between the two numbers (“about three thousand five hundred souls”), even though there is some chance that the original text had the and between three thousand and five hundred.