This is not quite accurate. Many COVID-19 patients are diagnosed before they are critically ill. Indeed about 80-85% of diagnosed patients are sent home to recover. A confirmed case is someone who has sufficient symptoms to seek, and ultimately have confirmed, a diagnosis. The CFR (case fatality rate) is calculated using this population as a basis. It is a remarkably consistent population of individuals. If you make it to the symptoms/diagnosis stage, you are in a known statistical bin with defined hospitalization and fatality rates. The CFR is quite sufficient, and highly useful, for planning for hospitalization and death rates, etc. Those that don't present with symptoms, or have mild symptoms, don't need treatment, and aren't going to die. The CFR is the standard for planning for virtually all diseases, in part because it is a readily measurable and reproducible statistic. For new diseases, it is likely to be the only useful statistic.
The IFR (infection fatality rate) is more useful to understanding the epidemiology of spread and achievement of herd immunity. Asymptomatic individuals can spread disease, not just the diagnosed cases. This is one important reason why we need widespread testing, as it affects or modeling of the current epidemic, and whether or not future epidemics are possible based on transmissibility in the general population. But we do have three pretty good studies of closed populations that had nearly complete testing: the Diamond Princess cruise ship passengers, the residents of Vo, Italy, and another small village in Iceland. In each of these studies (about 3000+ individuals each across a wide range of age groups), about 50% of infected individuals were asymptomatic. That is why the estimates of the CFR is about double the IFR, currently.
Think of it this way. If you somehow magically knew you were infected, you are in the IFR statistical bin, with a 1% fatality probability (averaged over all ages). You may or may not exhibit symptoms if you are in this bin. If you have enough symptoms to seek diagnosis, you just graduated to the CFR bin, with a 2% fatality probability (averaged over all ages.) It's just a statistical bin. Both bins are useful to scientists and planners. It's not an either/or or which is "better." Each bin has different significance, measurability, and use. You don't need to have perfect information to make good decisions. In most cases, you can NEVER get perfect information. You need to know the limitations of, and how to use, the good information that you have.