Based on your recent data extracts, you appear to be using the 5-year 2012 ACS sample. I tried to replicate your issue in STATA, but I do not see a PUMA code of 3300 associated with the state of Massachusetts (STATEFIP=25) in your data file. If you are still seeing a PUMA code of 3300 in Massachusetts, could you share the STATA command (or whichever program you are using) that you are using to generate this issue?
Keep in mind that the 2012 ACS sample uses the 2010 Census PUMA codes, but samples prior to 2012 use older PUMA codes. PUMA codes change over time. In fact, the 1990 Census PUMA codesdo contain a PUMA value of 3300 in Massachusetts. Is it possible that you have merged your 2012 ACS data extract to any other data files that may be using an older version of PUMA?