One of the hardest thing in software development happens already before any code is being written. Estimating how much development time some feature might take is, at it's best, a guesswork.
No matter how generous estimates are being made they are still almost always too optimistic. Even after the original developers estimate is doubled. Twice. On the other hand, on those rare occasions when the estimates are being beat they are being beated big time. In a longer project these are almost enough to compensate all the underestimates.
So while it's all just guesstimates the total amount adds up to a roughly accurate estimate. It's also much due to the fact that any work will fill the time allocated to it. It's ready when the time runs out. At least in the agile world. Some things are just pushed out to meet those estimates. We deliver the work we can within the time we estimated that the work will take. Without defining the work itself too specifically.
In a perfect world those estimates would be based on perfect information. But it would be impossible to even start estimating how long the work to collect that information would take.