From having lunch with guys from the mechanical engineering department (they were right next door at uni) my anecdotal second-hand wisdom is: Most larger engineering projects are as broken, tedious, expensive and unpredictable as most software projects. You will always find unsteady ground that nobody told you about, new last-minute requirements, shitty subcontractors making a mess and substandard parts that happened to be cheapest on the tender. From what I've heard, around 2 in 3 projects are over budget or over time in a significant way because something was massively screwed up somewhere and needed to be hotfixed, patched, rebuilt or -my favourite- reevaluated to be statically sound without structural changes.

I think the common software engineering myth about "real engineering" being somehow better and something one should aspire to is just really a myth.