Amongst those convicted of homicide, it is nowhere near that. Estimates place it at 20-30%, and of those, the highest proportion is for those convicted of homicide during a felony while others such as those who where convicted of DV-related homicide recidivate at closer to 10% (which makes a certain macabre sense.)
see eg http://nj.gov/corrections/pdf/REU/Recidivism_Among_Homicide_...
I meant the phrase "tiny fraction" in light of the discussion of T1 and T2 errors: no matter what rule is made, unless you incarcerate all felons until death, some fraction will recidivate. How to correctly weigh type 1 errors is difficult.
In the US the recidivism rate is over 60%. That's a gigantic fraction.