Block Groups with Unweighted Sample of 0 Persons

We combine 5 year ACS data at the blockgroup level to build neighborhood level statistics for our community - Cambridge, MA. 

While reviewing the 2013-17 data we found that two blockgroups have an unweighted sample population of 0 in table B00001.  One blockgroups has a very small group quarters population and evidently has a history of similar values.  The other blockgroup has a large group quarters population and a small household population.  In the past the second blockgroup has had an unweighted sample population of around 250 persons.  When we look at various blockgroup tables we do find values reported for both these blockgroups in the current iteration of the 5 year tables. 

Our question is - how are these values being estimated if the unweighted population upon which they are presumably based at least in part is 0?  Any light that can be shed here is appreciated.