[M T]

 the fatality rate is for diagnosed cases. There’s good reason to believe not all cases of mild disease are diagnosed or confirmed.

Yes! And we don't know if there are significant numbers of asymptomatic cases, because only symptomatic people are being tested! But viral infections in general do produce asymptomatic cases, no reason to think this virus will be any different. The denominator is unknown and unknowable at this point. The publicized fatality rates are estimates, not reliable facts, and are likely too high.

WHO's sitrep 30 (19 Feb) states

The confirmed case fatality ratio, or CFR,is the total number of deaths divided by the total number of confirmed cases at one point in time. Within China, the confirmed CFR, as reported by the Chinese Center for Disease Control and Prevention,9is 2.3%. This is based on 1023 deaths amongst 44 415 laboratory-confirmed cases as of 11 February. This CFR does not include the number of more mild infections that may be missed from current surveillance,which has largely focused on patients with pneumonia requiring hospitalization; nor does it account for the fact that recently confirmed cases may yet develop severe disease, and some may die. As the outbrak continues, the confirmed CFR may change

WHO's sitrep 38 (27 Feb) gives numbers that indicate current values of 4.0% CFR for Hubei and 3.5% CFR for China.
Hubei province: 2641 deaths in 65596 confirmed cases
China (total): 2747 deaths in 78630 confirmed cases.

The larger CFR now compared to 8 days earlier could be due just the maturity of the disease in the area.  Hubei has had more sick people longer, and more have eventually died. Other areas have not had many people sick for as long, potentially meaning that in a few weeks their CFR numbers will look similar. 

If, for whatever purpose, you want to make some other estimate different than CFR, you're free to do so.  But, the 4.0%, 3.5%, and 2.3% are measures, not estimates.

WHO-China Joint Mission Press Conference and article

A big question that still remains unanswered is how many mild cases go undetected, a metric that is key for getting a better handle on the disease's severity and fatality rate. Aylward said the group saw bits of information, such as testing of samples from China's flu surveillance system and testing of convenience samples, that hint the number of milder cases may not be great.

There are more interesting bits in that article.

Regarding the clinical spectrum, Aylward said about 88% of patients remember having a fever and 68% had a dry cough. Only about 12% had a runny nose, hinting that COVID-19 might not have much upper respiratory involvement. Some patients reported a prodrome that consisted of fatigue, muscle pain, and a vague feeling of being unwell.

[M T]

 the fatality rate is for diagnosed cases. There’s good reason to believe not all cases of mild disease are diagnosed or confirmed.

Yes! And we don't know if there are significant numbers of asymptomatic cases, because only symptomatic people are being tested! But viral infections in general do produce asymptomatic cases, no reason to think this virus will be any different. The denominator is unknown and unknowable at this point. The publicized fatality rates are estimates, not reliable facts, and are likely too high.

WHO's sitrep 30 (19 Feb) states

The confirmed case fatality ratio, or CFR,is the total number of deaths divided by the total number of confirmed cases at one point in time. Within China, the confirmed CFR, as reported by the Chinese Center for Disease Control and Prevention,9is 2.3%. This is based on 1023 deaths amongst 44 415 laboratory-confirmed cases as of 11 February. This CFR does not include the number of more mild infections that may be missed from current surveillance,which has largely focused on patients with pneumonia requiring hospitalization; nor does it account for the fact that recently confirmed cases may yet develop severe disease, and some may die. As the outbrak continues, the confirmed CFR may change

WHO's sitrep 38 (27 Feb) gives numbers that indicate current values of 4.0% CFR for Hubei and 3.5% CFR for China.
Hubei province: 2641 deaths in 65596 confirmed cases
China (total): 2747 deaths in 78630 confirmed cases.

The larger CFR now compared to 8 days earlier could be due just the maturity of the disease in the area.  Hubei has had more sick people longer, and more have eventually died. Other areas have not had many people sick for as long, potentially meaning that in a few weeks their CFR numbers will look similar. 

If, for whatever purpose, you want to make some other estimate different than CFR, you're free to do so.  But, the 4.0%, 3.5%, and 2.3% are measures, not estimates.

WHO-China Joint Mission Press Conference and article

A big question that still remains unanswered is how many mild cases go undetected, a metric that is key for getting a better handle on the disease's severity and fatality rate. Aylward said the group saw bits of information, such as testing of samples from China's flu surveillance system and testing of convenience samples, that hint the number of milder cases may not be great.

There are more interesting bits in that article.

Regarding the clinical spectrum, Aylward said about 88% of patients remember having a fever and 68% had a dry cough. Only about 12% had a runny nose, hinting that COVID-19 might not have much upper respiratory involvement. Some patients reported a prodrome that consisted of fatigue, muscle pain, and a vague feeling of being unwell.