The AGC beats IBM...In theory only!

Unlike the computer of the lunar module, the so called AGC, the IBM computer of the Saturn rocket was a serious computer, which could actually work.
Of course, it had the performances of its time, and could not be compared with the current computers both in terms of power and memory, but it was perfectly suited for the relatively simple work it had to do (there was not even even a human interface).
I have seen many documentations of computers, and the one of the computer of the Saturn rocket has looked perfectly normal to me, nothing suspect, unlike the one of the AGC which was the weirdest I have ever seen, with instructions which should not even existed (like the so called "hidden" instructions), unpractical instructions, a poor set of branch instructions...
NASA has produced a document in which the engineers have shown a bench of tests comparing the IBM computer of the Saturn rocket with the AGC, and supposing to show that the AGC was superior to the AGC...The only problem is that all these tests are complete meaningless jokes.
How could a computer which could not even have added one and one have been superior to a normal computer which was actually working?
if the Saturn rocket had been controlled by the AGC instead of the IBM computer, there is a complete certitude that it would have crashed shortly have having lifted off.
https://www.youtube.com/watch?v=cWMdIJKxHCI
Link to the technical document describing this critique:
http://www.ibiblio.org/apollo/Documents/CritiqueOfIBMApolloStudyReport.pdf