I've also had issues replicating ODE models from papers in the past. I assume you've tried contacting the authors to ask for clarification and haven't had a reply? If so, I think you should submit all 3, because without any original code available, your effort is likely to be representative of anyone else's. (And it's easily vettable as it'll be on GitHub). The onus is then on the original author to point out how (if at all) your implementation differs from theirs.