Yeah you are right and know a hell of lot more than me in this area but I think you have to allow the 'field of dreams' business model (i.e build it and they will come) to run free particularly in these bleeding edge areas. Its the same with robotics and I guess artificial intelligence.the geeks sometimes need the reality check though - for too long they have designed capabilities looking for a platform and not the other way around - ie a capability around a future identified requirement. however, they do an amazing job, DSTO geeks have made advances in acoustic warfare that the americans (eg) didn't think possible.
The market and CONOPS isn't always their waiting for you from the beginning. That doesn't render the research any less important.
Look at what Colin Angle has achieved. If he didn't blindly sink his teeth into robotics in 1990 (Post MIT) he would never of been in a position to pounce on the IED market which boomed in Iraq in 2002.
You have to let the geeks run riot but then you need this other group of "connectors" who have the strategic mindset to marry innovation to asymmetric threats. We have to always be ready to re-write CONOPS.
Sub design is obviously a very different ball park to little plug and play 'modular' weapons platforms in Iraq.