Can I compare the 2016-2020 5-year ACS to the 2021 1-year ACS since we’ve been cautioned against using the 2020 1-year experimental dataset?
Yes, of course you can, but is it meaningful? I would recommend comparing single-year to single-year estimates; and comparing five-year to non-overlapping five-year estimates. Definitely examine single year estimates 2005-2019, 2021 (skipping 2020!). Definitely compare 2011-2015 to 2016-2020 five-year estimates
I would compare the 2016-2020 five-year ACS estimates to the 2018 single-year estimates if my intent was to "adjust" the 2016-2020 small area (tract-level) data to simulate/represent my best estimate of 2018 "base year" estimates for demographic/transportation modeling analyses. I would generally consider the middle year of a five-year period to be the "fulcrum" year - - representing the "half-way" mark in the five-year period.