That's an interesting observation, but due to the "rank modification for spam detection" ( [webmasterworld.com...] ) and the "near immediate change in position" after the removal of the <b> elements I find it difficult to draw a conclusion from it, because when you change "factors that impact rankings" your results "go for a random roller-coaster ride" over the next N (randomly generated to the best of my knowledge) period of time.
I would guess adding a new <section> (as would be determined algorithmically by the H3 addition even if not explicitly sectioned in the HTML) would be considered a "ranking impacting factor" so the results you're seeing could easily have been the same if you had left the page completely alone and "let the random ranking roller-coaster run it's course".
Also, from what I've seen lately, even if a page is spidered the day of or the day after a change is made seeing the difference in the SERPs is taking up to a week to "filter through the system" and get into the "regular" SERPs whether the page is updated for a site: search or not. (I've tested this on a large number of pages over the last two months and there's been a definite "lag" between spidering, slightly new version found, site: search update and then the SERPs showing the new version of the page.)
I'd be interested to hear if you're 100% positive the page was respidered without the <b> elements after they were removed and the version without the <b> is 100% for sure the version in the SERPs rather than the original you launched with the <b> elements in place.