When you consider the limited tools NASA had to work with, the shuttle was one of the greatest aerospace achievements in history. Besides all the issues with structures, dynamics and propulsion on ascent, no winged hypersonic vehicle (manned or unmanned) had ever reentered from orbit. There was very little data.
When the shuttle was designed in the mid-1970s, CFD was incredibly primitive, and hypersonic wind tunnels extremely limited. Yet the shuttle successfully flew a manned mission on the first flight.
Decades later, one reason (of many) NASP failed was continuing limitations in CFD and hypersonic wind tunnels. Unfortunately this is still a problem today, despite the much faster computers. This is due to limitations in CFD models and the fact that most demanding CFD problems are four dimensional. Every factor of two improvement in grid resolution requires a computational speed increase of 16 times. This is why hypersonic test vehicles like the X-43 and X-51 are so vital: even today, CFD cannot comprehensively simulate the required flow characteristics, and hypersonic wind tunnels are too limited in speed, duration and data quality.
This history is detailed in the book "Facing the Heat Barrier: A History of Hypersonics", which is available on line in PDF format:
Part 1:
history.nasa.gov/sp4232-part1.pdf Part 2:
history.nasa.gov/sp4232-part2.pdf Part 3:
history.nasa.gov/sp4232-part3.pdf
SpaceX is doing a lot of leading edge work in CFD. Here is an interesting article with some links to some simulations:
Rockets Shake And Rattle, So SpaceX Rolls Homegrown CFD