The media and public have been asking questions on how the modeling/predictions from "experts" could be so wrong. Original numbers from models were much, much higher than what we are actually seeing.
I've worked as an engineer in the oil industry, all my adult life. Oil companies do lots of modeling or what is usually referred to as simulation. We have a saying, "All models are wrong, however some are useful." Models are good for predicting the direction data will move, but it's good to not hang your hat on the actual outcome on early models.
Models are based on a lot of assumptions. The more you know about what you are modeling, the better assumptions the modeler can make and the better results or output from the model will occur. In the oil industry, we will model reservoirs or entire oilfields with producing lives of 10, 20, 30 or more years. As you collect real data, the model can be re-calibrated until the model gets a good history match of what's actually occurred and then give a better prediction on future performance. Theoretically, a better prediction is made each time you re-calibrate with more real data. People that do the actual modeling fall in love with their models and fall into the trap of believing them to be gospel.
Modeling the COVID-19 is problematic due to lack of knowledge on the subject. For instance, if you look at something as simple as how quickly will the number of cases or the number of fatalities double. This has been what the media has focused on, asking "how can the numbers be so wrong?" If you do some simple math, say compare a doubling rate of 4 days vs 5 days and run the model for 60 days, you can clearly see how the numbers diverge substantially. If you make the assumption that something doubles every 4 days, you go from 1 on day 1 to 32,768 on day 60. Changing the doubling rate from 4 days to 5 days, you go from 1 on day 1 to only 4,096 on day 60. So you can see by this example that a seemingly small change to a key assumption makes for a large change on the prediction.