This is an estimate with a whole bunch of assumptions (may or may not be true) and known factors and applied to a probabilistic formula (a very simple one too.) Some I'd say this is a very rough attempt.
The 440 confirmed cases released by the daily official press release is a fact, with all of them confirmed with lab tests and been quarantined. Those are live numbers: https://3w.huanqiu.com/a/c36dc8/9CaKrnKp248?agt=8
There is no reason to believe the simply probabilistic model would be accurate (this is also why they suggested that the accuracy is between 1000-9700, almost an order of magnitude difference). And there is no reason to believe their assumptions were right either.
So you are comparing Apples to Oranges.
The best modeling for estimates about infectious disease is to use ordinary differential equation (ODE) to model and estimate with a set of assumptions. It talks about why the ICL's model only applicable to the early stage of the outbreak.[1]
[1]: https://mp.weixin.qq.com/s/Ab0fLOWr_9AvU4RA15rMCg (in Chinese)