But once established they definitely used patents against smaller competitors. So it is hard to say for sure even in this case whether patents were a net win. They probably were though. The patent probably caused modern separate-condenser steam engines to happen at least a few years earlier.
(I recommend the story of Bolton & Watt to present day founders. It is remarkable how little has changed.)
One reason is independent discovery: most innovations tend to be discovered independently and nearly simultaneously. Like, when "science is ready", innovations just pop out of several people's mind relatively quickly. (See also Eben Moglen's folk theory of the internet and innovation: "wrap the internet around a planet, spin the planet, and ideas flow out of the network".)
If we accept that, the only thing that could have significantly slowed the development of the modern separate-condenser steam engine down is money. Like, no single one investor is willing to risk making the first step, by fear of being eaten by copycats. If this is so, then we have a tragedy of the commons.
Risk aversion in this case is indeed a problem. I think it is mitigated however by (i) the first mover advantage, and (ii) variations of risk-aversion among individuals (combined with independent discovery, you get a winner-takes-all scenario).
Ah, before I forget: adopting an individualist point of view tend to favour patents: an inventor is entitled to reward (forgetting other's entitlement for freedom of invention). If I didn't have that patent, then I wouldn't have done this much good (forgetting that others may have done it otherwise). I know you didn't make this error, —Cf your §2—, but others do —for instance by taking your §1 out of context.)
[1]: http://www.dklevine.com/general/intellectual/against.htm
I think the key takeaway is that the patent was retroactively extended; so the idea was already created. If the extension was necessary for funding and development, then a case can be made for granting temporary monopolies for anything regardless of the inventor (the idea already exists, so we should find the best implementer) - I think most would be against this idea.