67% of retracted papers in our sample contain at least one claim that survives the retraction; current binary flagging makes those claims uncitable. Replication status: untested.
lines 89–94 in main.tex