Why are some observations classified as SPM poor even though SPM resources are greater than SPM poverty threshold?

In 2016 ASEC (but not 2012-2015 surveys) some observations with SPMPOV=1 have SPMTOTRES>SPMTHRESH. This doesn’t seem to be a rounding issue because in one case total resources are almost 5% above the threshold. What am I doing wrong?

You are not doing anything wrong. Actually the Census Bureau has recently released a new version of the SPM variables for 2016, which fixes the issues you highlight. These revisions will be reflected in the next IPUMS CPS data release.