It's common for data to be suppressed when there isn't a sufficient sample size. The suppression takes place to protect information about the respondents in those areas and to limit unreliable statistics. In many cases there are limited numbers of people who have a disability or are within the age group of interest at the block group level. The counties have a larger population so you likely won't get null values at that spatial resolution.
The best work around is to use 5-year estimates rather than 1-year estimates because they aren't subject to the same data suppression guidelines. If you are still running into this problem with 5-year estimates you will have to get creative but it's hard to say what your options are without more information.
Below is a link to the data suppression page for ACS:
www.census.gov/.../data-suppression.html