I guess this is a feature request?
Seems like preselecting MULTYEAR when selecting a PUMA for a multi-year ACS sample is a good idea. Otherwise the PUMA is really difficult to interpret. For example, for a 5-year ACS sample spanning 2010-2014, the 2010 and 2011 observations use the 2000 PUMA convention and the 2012, 2013, and 2014 observations use the 2010 PUMA convention. Without the manually selecting the MULTYEAR variable, you cannot determine which PUMA geography to use.
The workaround I’ve used - without creating a new extract, downloading, unzipping, etc - is to match the 2010 PUMAs to the entire dataset, then match the 2000 PUMAs that aren’t in the 2010 PUMAs to the remainder of the dataset.