They're just saying that building it is not a proof - the computer simulation is more exact than the model, and even the simulation is subject to rounding error

They're just saying that building it is not a proof - the computer simulation is more exact than the model, and even the simulation is subject to rounding errors in what is considered "touching". Sounds like they used the computer to search for a proposed solution that _could_ have been a false positive due to rounding error, then verified it through exact mathematical methods.