Never said anything else. The rows I've mentioned were clearly the rows of bits (represented by presence or absence of a hole). From a pure theoretical point of view it still appears to be wasteful (as the used character set encodings didn't use all representations of 12 bits), but there were "good enough" reasons behind that decision too: It all started from the design of the tabulating punched card from the 19th century, and from every digit 0-9 having its own row, and two more rows for other marks. The hardware depended on these properties.

125 years ago, the logic I've described was already there, 12 rows but still less columns on this card:

https://en.wikipedia.org/wiki/File:Hollerith_Punched_Card.jp...

Increasing the number of columns was obviously later technical improvement.