In the reactivity measurements of the China Experimental Fast Reactor (CEFR) start-up tests, five independent control rods were moved to compensate for the reactivity insertion. Since the recorded control rod positions were converted to reactivity via S-curves (rod worth curves), any uncertainty in the S-curves can propagate to all reactivity measurements. In this work, we rigorously derived the uncertainty of the reactivity in terms of the statistical uncertainty of the S-curves with Monte Carlo (MC) simulations. Additionally, the average error of the control rod worth from the MC calculation was estimated from experiments and embedded into the uncertainty formulation. The formulation shows that the error of the reactivity is highly correlated to the changes in the control rod position during the reactivity measurements. McCARD MC simulations were then conducted for the CEFR start-up tests, and the calculated reactivity and uncertainty were compared with the measurements. The main error factor of each reactivity calculation was figured out by quantifying the uncertainty components. With the uncertainty formulation, the calculation results showed a better agreement compared with the measurements, as the relative errors were observed mostly within 2σ of the uncertainty.