While time is down, it's the time of a hypothetical user entering the command, not the time at which commands actually affect data in the table. I think that's pretty intuitively clear to the reader, especially since the article is about counterintuitive behavior.