The Middle Indo-Aryan of the Ashokan edicts had lost the syllabic vowels entirely. The sound fell together with a, i or u depending on the dialect, which is one of the differences by which the dialects of the edicts are told apart, and a scribe of the third century BCE writing Prakrit therefore had no occasion to cut this letter at all. It belongs to the Brahmi that was later adapted for writing Sanskrit, when the script came to be used for a language older than the one it had been developed to record.
That gap is worth dwelling on, because it says something about how the alphabet is transmitted. The letter is in the order, and the order was recited whole, so the sign was carried forward by a tradition of memorised sequence rather than by continuous use on stone. When Sanskrit began to be written in the early centuries of the common era the letter was there waiting, in seventh place, ready for a sound that the writing had never had to record. Every script descended from Brahmi has it, including several used for languages that have no such sound and never did. A place in a memorised order is a durable thing, and it does not depend on there being a word that needs the sign.
The letter is nonetheless among the more common in later Sanskrit epigraphy, because the roots and stems it appears in are ordinary ones and because a great many personal names are built on them. So its attestation is thin in the third century BCE and thick from a few centuries later, and a palaeographer dating a hand from it has to be careful which end of the sequence is being used as the standard. That uneven distribution across periods is a general feature of the Brahmi evidence rather than a peculiarity of this one sign.