OTOH, the cost of JWST has blown out even further than Hubble (approx $9b, from an initial budget below $2b) precisely because there's no human servicing, which means everything in the overly-complex design must deploy perfectly or the entire mission is a bust. Eliminating the added cost of making the spacecraft serviceable is more than made up for by making the need to ensure the spacecraft can't fail.
So "the science guys" aren't a guarantee of savings, once a robotic mission becomes the flagship program and everyone tries to latch on to the teat to fund their idiotic ideas.
The problem with HSF at NASA is the legacy of Apollo, the hundred thousand employees and contractors, the scattered NASA centres and even more scattered contractor networks, which all make HSF unaffordable. (For example, the annual cost of the Shuttle program was the same regardless of how many missions they flew that year, 6, 4, 2 or none. The annual budget for operating the completed ISS is, by amazing coincidence, exactly the same as the annual budget during the construction, which was by yet another amazing coincidence, exactly the same as the annual budget during the last four years of development.)
By developing private human space-flight, we can reduce the cost of doing on-orbit repairs until it's cheaper to send humans to fix something than to write off the spacecraft and send up a new one.