Number of bedrooms exceeds number of rooms

I've found a data issue while using the ACS 2010-2018 1-year microdata, accessed through IPUMS. In a small subset of the data, sometimes "bedrooms" exceeds "rooms". I contacted both IPUMS and the Census, and neither were able to figure out why the issue exists. Has anyone else encountered this issue or does anyone understand why it might exist? I've been able to narrow down that it exists only for 2010-2014 and for households with 6 or more rooms.