I'd soften "irrelevant" a little. For US CPI the figure that does most of the work in rate markets is core month on month, and the first move usually follows it rather than the headline. Knowing which figure the market is watching that month helps you read the reaction rather than just wait for it.
On the journaling, the sample size is the thing to plan for. There are only twelve CPI releases a year, so a cpi_week tag might collect a few dozen trades over several months. An expectancy figure from that many trades has a wide margin of error, and one bad week can make it look negative. I'd let it run for a full year before changing the rules on it, and compare against the non-CPI weeks over the same period, since a slow month can affect both.
CPI weeks also tend to bring PPI and retail sales within a few days, so the tag effectively captures a cluster of releases rather than just one. That's not a problem, as long as the conclusions take it into account.