At a direct cost of around 1.5 billion dollars per flight. This includes a lot of flights that were essentially milk runs, bringing satellites into orbit or doing other light work that could have been just as easily done with much cheaper capsule flights.

There was a lot of wishful thinking with regards to the shuttle program. Some ridiculously fanciful, some plain ignorant.

But most of all, it was the test of a theory put into practice. And it should have been to no one's surprise that many things thought true wern't. How much of that was simply ignorance vs being deliberately oversold is a matter of debate. But what those 133 missions did was impart an operational knowhow and lessons learned that went far beyond the immediate mission objectives. These are the same benefits provided by the ISS today.

In short, you really can't begin to determine how well a system works in practice until you put it into production, and start using it a lot. These are the same things SpaceX is learning now (just of a new sort.) It's why Ares is returning to a capsule form.

Buran looks much better on paper, and I've no doubt it's an engineer achievement. But the success of a program, and the benefits derived from it can't be determined purely on a theoretical basis. Buran may have been an excellent orbiter, or it could have been terrible. But we'll never know.

A large number of those missions were flow for the DoD, repairing/refuling satellites and potentially also capturing enemy satellites.

Large? Of the 135 flights, almost 50 went to Mir/ISS, doing work that should have been done by Soyuz/Progress/TKS-equivalent spacecraft in the $100-200 million range. 32 more made Spacelab flights, aka "poor man's excuse for a space station". The remainder were split between commercial satellite deployments (best left to unmanned boosters), regular scientific missions (the same already done on Gemini and Apollo capsules), missions that could have been cheaper put on a satellite (e.g. radar mapping), and very few missions the Shuttle was actually useful for – DoD missions and scientific recovery of large payloads (the only mission profile the Shuttle alone was uniquely suited for). Even Hubble's repair could have been done with capsules (the JWST is planned to be maintainable by Orion capsules, should maintenance ever be needed).

There were at least 11 classified NRO missions, and likely more with NRO tasks. Requirements from the NRO directly influenced the shuttle program design. I guess "large" is an exaggeration though.

They obviously didn't use Shuttle exclusively for military purposes. But most people don't realize that nearly all government-funded scientific programs in the US also have significant defense/NRO objectives. Even The Dish at Stanford was funded to be used for SIGINT work during the Cold War.

I'd wager that these groups are now using advanced robotics to replace classified work previously done via Shuttle missions. i.e. Boeing X-37

I don't think there's any maintenance plan for JWSP, capsule or otherwise. L2 is way too far away. Do you mean maintenance after launch but before deployment?