If you check the intellience report from the ODNI, they lay out a little bit where these differences come from. For examples, some agencies like the FBI do not give much weight (or any) to science, and rather rely on human intel as well as interpretations of Chinese obfuscation.

There might also be some political interference, since we know that a lot of the human "intel" is from like memos and other garbage conjured up by ex-Trump state department people like David Asher.

Third, there is a selection bias for "natsec" types that judge based on their institutional mission towards preventing existential risks and bioterrorism, now getting a windfall and more resources because of the controversy; which might again make them more amendable to disfavor and discard scientific evidence and keeping all options on the table.

Lastly, this is a phenomenally complex topic and most of the science, even in this article, is not as fast as the news cycle. I assume that with the integration of some of the (still preprints) highlighted here, there will be a shift from some agencies in the upcoming year/years on this issue towards natural emergence as well.