Sure, estimates are, well, estimated... based on what the person estimating knows at a point in time, the choices are (1) I estimate it it will require "X" to do it, or (2) I can't estimate what is required to do it. I think this comes down to people not wanting to accept that an estimate could be wrong, or not right enough.
I get an estimate when I need car repairs, to make sure I can or want to afford it. My mechanic calls me when he finds something he did't know that will change the estimate, and I decide what to do. If I find over time that my mechanic always under-estimates the cost, eventually I will change mechanics. The average reality should be a bell curve around an estimated value, an estimate is almost never exact, but a statistically acceptable variation is what you should look for.
Now, such analogies of physical things are problematic when used to describe an aspect of software development. Software is intangible for one thing, and we really haven't been creating it very long to have some good tools like other disciplines (like Engineering) for predicting outcomes; but like everything else, it takes time and money to create it, so the people investing that time and money usually want an idea of how much that investment will be to get what they need. Just saying an estimate is worthless isn't really going to go over well.
The advantage of software over physical things is that you can indeed create small amounts of it that can be seen to work. Long before 'agile' got its name, a lot of software projects switched from estimates to time-boxing: give me x amount of resources and the project will deliver what it can in that 'box'. Investor spends a little to see what he can get for that amount. If it looks good, run another time-box. If it doesn't look good, you can stop with only a small investment amount having been spent.
Lastly I would comment on your statement on what does an estimate have to do with creating good software and a good return on investment. ROI is a funny thing, often very date-specific; delivering great software after a competitor has already done it and captured the market will kill your ROI. Some windows of opportunity are very short; miss them and the return is nil. In these cases, you have to have some way of knowing if you can make the window. Like it or not, an estimate is one way of trying to figure that out. It is these kinds of situations that will make investors wary of new development and look for packages or services to meet the need. Nothing is perfect but not being to estimate software development may just lead to no new software being developed.