"So due to the transaction costs, it is not better for society"
It seems you are implying that when the risk is in the side of the producer there're not costs for society.
That's not my view at all; I'm sorry that it came off that way. The part that was implied and probably shouldn't have been is that a more closely estimated risk has a significantly lower cost than one that's less closely estimated. This is a common economic theory; it is generally considered that higher risk (variation) goes with a need for higher returns (see e.g. modern portfolio theory).
I also think there's several reasons why the risk will be lower if taken by the software developer rather than through contracts for future payment. As a software developer, I can change the scope of the project. I can start out with something that I think will be a Windows program, and then if Mac becomes popular among my target group, I can switch to having a user interface for Mac (only) instead. Or I can switch to a web interface. Or I can switch my product around, and instead of having it target general music enthusiasts, I can find that my beta DJ users love it and would be willing to pay a significant price for it, while the general music lover says "Meh." In the worst case, the developer can cancel it if it takes too long to develop. However, if I'm going to pay for somebody to develop something, I am not willing to pay for "We'll deliver it on some platform; and it will do something that some people are interested in." I want a specific set of features, for a specific platform. This adds risks for the developer, as they lose flexibility.
There's also the important part of time delayed delivery and customer circumstances changing. If I have something that takes a while to develop, it's likely that some of the potential customers have moved on to other things, while some new customers have come along. As an extreme example, if I'm developing a game that target 12-13 year old girls and it takes four years to develop, *all* potential customers will have moved on and been replaced by new customers. As a less extreme example, I might have switched to wanting my software on the Mac or on the web or on Android - while having paid for what seemed reasonable at the time. This adds risks for the customer.
Bad news: the costs are exactly the same since a failed project is a failed project anyway.
But the costs of a failed project are different, and the value of the money is different. As an example, if I'm in a startup, software that I can use now has a high value, and cash in hand has high value - while software that I get delivered in a while has much less value.
But the basis of marketplace-like capitalism is that both producers and consumers are perfectly informed of their options
I think you're thinking of what defines a perfect market / perfect competition (http://en.wikipedia.org/wiki/Perfect_competition)?
which is less the case when there are production efforts that are invisible for the consumer. Add to this that the "trick" for license-based business is that while the production costs are bounded, benefits are not. That's what allowed to the owners of the software giants from the eighties-nineties to become some of the richest people of the world in record time spans. Whenever you see net benefits going well over 100% you can bet capitalism is not working as expected and society as a whole has a worse deal.
You're thinking of "increasing return to scale". That's certainly one of the violations, but there's also violation of "Homogeneous products", "Perfect information", and probably some others.
In the specific case of Microsoft, they manipulated the market by tying their product to another product (CPUs), making sure that if anybody else wanted to sell a competing product, that product would have to be bough *in addition to* Microsoft's product.
As for net benefits going well over 100%: It really depends on what risk is being taken. If you have 100 companies that start up in an area, and you expect 99 of them to fail and the group as a whole to have neutral return, you expect over 100x return on the last one (100x + inflation over the time period). If you afterward look at only the single one, it'll look like a capitalism failure, but it's really flat risk handling.