Estimating in highly inaccurate resolution, eg fibonacci or powers of 2, is a way to embed the lack of confidence in single number estimates.
You may think something takes 9 days, but if you are only allowed to answer either 8 or 16, choosing 16 is the only safe option and automatically includes some lack of confidence.
While that's true, it doesn't work if everyone doesn't have the exact same understanding of how it works. For example, most people I've worked with will choose 8 instead of 16 for an estimate of 9. This makes it risky to overestimate and appear less capable.
It's a suitable approach in the local context (tickets/tasks for the team), but won't work on the project level and with stakeholders. Stakeholders eventually will want a timeframe.
You may think something takes 9 days, but if you are only allowed to answer either 8 or 16, choosing 16 is the only safe option and automatically includes some lack of confidence.