> The only correct estimation is actually doing the work and looking at how long it took.

You seem to be missing the point of 'estimation' (as in, it is an approximate prediction, not an accurate measure). Admittedly management also tends to miss this same point, and that fact puts a lot of us on the defensive - but the answer can't be to throw estimation out entirely.

If you ask a civil engineer to estimate the cost and timeline of building a bridge, they'll get you an estimate. It'll likely be within 1-2 orders of magnitude of the actual cost/time, due to a bunch of unknown-unknowns that are hard to factor in (things like: how long the legislature will take to permit the build). But it'll still give you useful data about the differences between building a $million foot bridge, and a $100 billion highway bridge...