Earlier works, including a recent one at BNL, demonstrated that positron emission tomography (PET) is a potentially powerful tool for quality assurance and the treatment planning of Proton Therapy (PT). In particular, the PET images show the adequacy of the overlapping of the dose distribution with the intended target volume. Here, we present calculations of the possible variations in yields of positron emitters produced by proton beams of 250 MeV in a soft tissue, and combine the results with the Monte Carlo simulations of PET data collection from our recent work to show the effect of input data variations on the images. We demonstrate that the image results depend strongly on the available nuclear reaction cross section data.. The emphasis of this work is on determining, quantitatively, the differences in the calculated PET image yields resulting from four different sets of input nuclear reaction cross section data.