HUMINT, IMINT, SIGINT, GEOINT, MASINT, are just some of the methods, and there are probably more. As for your claim US intelligence agencies are no better than rumors on public forums, that's a counterintuitive claim. Can you explain why you believe that? And do you think it's peculiar to US intelligence agencies or it's the case with all agencies, including the Communist Party intel services?
Keep in mind that ONI is not the entirety of US intelligence, or even the most informed of the US's various intelligence agencies. In the grand scheme of the US's intelligence establishment it's somewhat peripheral. Furthermore, all intelligence reports will involve some kind of speculation in the face of imperfect and incomplete information, so you can't take everything on a report at face value. They're not free of inferences which must be examined and assessed. That said, what we are disputing over isn't even an intelligence report. It's a white paper for public consumption. It's probably not even written by people who have intimate knowledge of the subjects they're only summarily and lightly covering.
Also keep in mind USCC reports aren't intelligence reports. I know people who know people who've worked for the commission, and they're not privy to any special information not available to the public. If the USCC white papers corroborate with the ONI white papers, it's because the former uses the latter as a citation (pay attention to their references).
That said, even putting the question of how useful ONI reports are aside, keep in mind that China frequently surprises the US with its military developments. That's for a reason. The US intelligence establishment is vast and capable, but it's not omniscient, and it's certainly not infallible. Countries like China, Iran, and NK are rather opaque to the US, especially since the lack of cultural affinity and familiarity creates barriers in HUMINT, and so much of good intelligence work leans on HUMINT being at least the starting point for other methods (I would be willing to wager a lot of money that Taiwan's intelligence community has a much better picture of the China's technological progress for these exact reasons). Furthermore, when trying to collect intelligence from large bodies, the same problems we have as hobby observers trying to figure out the reliability of different sources also comes into play for actual intelligence work. Even insider informants, intercepted signals, or hacked databases won't ensure the information you collect will be relevant or accurate. That person working in a cubicle in the DIA or CIA with reems of collected information may or may not have more to work with than we do, but they are, all the same, as beholden to the whims of inference and interpretation as we are. That's just the nature of intelligence work.