Situationally. With programming there are loads of different tasks. For some tasks I think time can be quite a good metric. Still, metrics matter and I think having lousy metrics is better than not having metrics at all. Of course you have to realize what purpose the metric has in certain situation.
I fundamentally disagree. Metrics don't live in their own world. They affect the world around them and incentivize certain actions. Bad metrics can produce bad incentives. There are certainly many cases where simply not knowing something is better than knowing something that's only partially right. However, that requires a level of humility and acceptance of one's limitations that can be hard to defend to others looking for straight answers to all questions.
But generally I think my main point stands. We need to be humble enough to sometimes accept lack of control (i.e. not using certain metrics/processes) rather than deluding ourselves by through certain metrics that really aren't very good.
Never mind that the task spec changed 5 times in those 3 days.
Because: 1) you know what you are lacking 2) you won't take action on wrong information.
But if you're a manager giving out raises to your team, taking no action (i.e. giving no raises) might seem a worse choice than using the unreliable data - and there's no expectation reliable data will arrive if you wait.