I realized that the number of microdata cases represented in each puma is lower than average for the missing PUMA2010s, but still curious why geocodes are included in microdata but not shapefile. Shapefile retreived from: https://usa.ipums.org/usa/volii/pumas…
The codes 3505-3519 do not appear in Illinois in the 2010 PUMA shapefile because those codes are used in Illinois only for 2000 PUMAs, not 2010 PUMAs.
ACS samples use 2000 PUMA codes through 2011. Multi-year samples (e.g., 2011-2015) include PUMA codes for both 2000 and 2010 PUMAs. The reported PUMA codes and definitions depend on the year each respondent was interviewed. To identify the year of interview (and by extension, which PUMA definitions are used), you could use the MULTIYEAR variable.
For more information about PUMA definitions, see the IPUMS USA Geographic Tools page.