It makes sense for the article to not distinguish between SIGINT and HUMINT. It's just about the effectiveness of misinformation. I guess the idea is that intel

It makes sense for the article to not distinguish between SIGINT and HUMINT. It's just about the effectiveness of misinformation. I guess the idea is that intelligence gathering is only useful in cases where counter-intelligence is difficult. The examples given in the article were cracking widely used codes (because it would be hard to fake large volumes of traffic) and plans for secret weapons (because they can be verified by building the weapon). Given the examples discussed here, I would guess that a single trusted embedded agent with a long history of providing valuable information could be trusted up to a certain point. I suppose anything that is easily checked would be good too. Bear in mind that coming up with an example where HUMINT was key to victory doesn't actually prove your point at all, because there may be even more examples where counter-intelligence was key to victory.