I believe it's all about size. Small tasks can be estimated fairly precisely, with 30% margin maybe. But some people extend this logic to huge projects and this simply doesn't work.
I give you example: I was once asked to estimate the amount of work needed to migrate a huge org from on-prem to one of the public clouds. Without even knowing what kinds of app they were running. I said I can only do a very rough estimate and there are too many variables to be precise. But they insisted and hired someone else to do this job with the precision of half-day!
Needless to say, their estimates were too optimistic because they could only include what they knew. By design, they couldn't measure the things they know nothing about because these issues haven't appeared yet. Last time I checked, they are still migrating and I have a feeling they will finish around the time I had originally planned, that is the second quarter of 2024.