> Repeating someone's experiment but changing several of the variables (temperature, overnight stop versus all-in-one-day) does not refute the experiment.

That depends. If the stress on the car's battery is the same, but the driver's behavior with respect to charging is different, then the new experiment may have important things to tell us.

And remember -- in science, there are no failed experiments. A replication effort that doesn't produce the same result as the original experiment is still a useful result.

Right, but not having the car parked in a lot overnight in very cold weather would seem to be a very different type of stress in the CNN run.

Yes, true, which means we have a new data set, and if we're careful in our analysis we may discover why the new data differs from the old. Therefore it's a successful experiment.