I tried to reproduce collapsing the 42 languages in table B16001 into the collapsed 12 languages in table C16001 and failed. After deep diving in the documentation, I found the cause.
The 42 and 4 language table that came with the recent detailed language release lists Telugu, Tamil and other Dravidian languages as Indo-European languages, and I found a similar table with the 2016 changes to the B16001 table; I assumed that in C16001 they were classified as Other Indo European languages but the numbers didn't add up.
Then I found documentation with the previous detailed language release (2019) and that grouped Dravidian languages with the Asian languages. This was the key to creating a walkover table between B16001 and C16001. The 42 and 4 language group tables I found are NOT reflecting how the Census Bureau processes the languages, In the collapsing process, the Dravidian languages are grouped with the Asian languages.
After I knew where to look, I also found in the 2016 language user note "All Dravidian languages were included under “Other Asian languages” in the 2015 B16001 table." without mentioning that they were reclassified as Indo-European, which would have been something worth mentioning.