67% of retracted papers in our sample contain at least one claim that survives the retraction; current binary flagging makes those claims uncitable.
lines 105–108 in main.tex