Incomplete Preferences in Choice Experiments: A note on avoidable noise and bias in welfare estimates