I spent some time poking around the CDC site and wasn't able to confidently confirm the source for the data, alas. My hunch is two possible reasons. First is data collection itself - it's possible the sources for the data changed or data collection methods were improved so the dip reflects a shift from estimated to actual. I can't access all the books the article cites but it's also possible the author of the graph was aggregating from different sources.

The second reason might be the shift from predominantly home births to hospital births. From the page hosting the graph:

Obstetrics as a specialty was shunned by many physicians, and obstetric care was provided by poorly trained or untrained medical practitioners. Most births occurred at home with the assistance of midwives or general practitioners. Inappropriate and excessive surgical and obstetric interventions (e.g., induction of labor, use of forceps, episiotomy, and cesarean deliveries) were common and increased during the 1920s. Deliveries, including some surgical interventions, were performed without following the principles of asepsis. As a result, 40% of maternal deaths were caused by sepsis (half following delivery and half associated with illegally induced abortion) with the remaining deaths primarily attributed to hemorrhage and toxemia.

While it wouldn't be until the 1930s that nearly all births were in the hospital, that dip could likely be attributed to the various unintended negative consequences that came with that shift.