Experimental yields, both independent and cumulative, of fission products from thermal-neutron fission of 235U, were combined with semi-empirical model values for those yields for which experimental data were lacking. Using weights determined from experimental errors, or from a priori estimates for model values (which weights were uniformly lower than those for experimental values), and imposing the constraint that cumulative yield is a sum of independent yields of precursors, a most likely consistent set of yields and their errors was determined. The errors were adjusted upward in all cases for which the inferred consistent yield differed by more than its error from the ENDF/B-IV value. Using these yield errors, the sensitivity of decay power to yield uncertainty was determined both for a fission pulse and for very long, low-flux irradiation.