Nuclear power is a controversial topic and everyone is biased, especially IAEA, but even WHO. The author of the book that claims that 985,000 people have been killed by Chernobyl is also biased. I too am biased. However, I have no association with the nuclear industry and I don't personally know anyone that died because of the Chernobyl disaster.
That being said, I feel like all the answers up until this point has been very one-sided. The number 985,000 is based on a scientific study and I found it peculiar that no one addressed the study presented in this book. In order to be able to criticize a number you need to know what it's based on, right?
I did some digging and actually found the book freely available as pdf:
Chernobyl: Consequences of the Catastrophe for People and the Environment
It's 349 pages, so I didn't read all of it and I certainly don't have time to make a full review. I found this Wikipedia article about the book where they present the following quotes from reviews of the book:
"The introduction addresses the issue of why assessments of health effects from Chernobyl are so disparate. The authors disparage the approach favoured by the majority of the epidemiology community, which seeks a correlation of health effects with levels of contamination or dose. They believe this approach is ‘impossible’ due to lack of measurements in the first few days, lack of information on ‘hot spots’ and lack of information on all of the isotopes involved."
"They consider that the USSR authorities distorted links between health effects and radiation exposure and they prefer therefore to rely on what they consider are independent investigations of comparative health measures in various territories that they consider are identical in terms of ethnic, social and economic characteristics and differ only in the exposure to radiation."
George Monibot comments on that review:
A devastating review in the journal Radiation Protection Dosimetry points out that the book achieves this figure by the remarkable method of assuming that all increased deaths from a wide range of diseases – including many which have no known association with radiation – were caused by the Chernobyl accident. There is no basis for this assumption, not least because screening in many countries improved dramatically after the disaster and, since 1986, there have been massive changes in the former eastern bloc. The study makes no attempt to correlate exposure to radiation with the incidence of disease.
However, his comment is disputed:
The fact of deliberately choosing control groups which they judge not to have been exposed for the purpose of comparison appears to contradict Monbiot's assertion that Yablokov simply assumed all increased deaths were caused by the Chernobyl accident. The review states that excess deaths recorded in the unexposed areas cited in the book were not alleged by Yablokov to be caused by the Chernobyl accident.
Another Wikipedia article stated that:
Between 1986 and 1992, it is thought between 600,000 and one million people participated in works around Chernobyl and their health was endangered due to radiation. Because of the dissolution of the USSR in the 1990s, evaluations about liquidators' health are difficult, since they come from various countries (mostly Ukraine, Belarus and Russia, but also other former Soviet republics). Furthermore, the government of Russia has never been keen on giving the true figures for the disaster, or even on making serious estimates. However, according to a study by Belarusian physicians, rate of cancers among this population is about four times greater than the rest of the population. All the figures quoted by various agencies are controversial
I included the last quote just to point out that things like increased cancer rates in these areas are likely due to the Chernobyl disaster. However, it's hard to prove that a particular individual got cancer because of Chernobyl and not due to other causes.
I think the main difference between WHO's report and Yablokov's is that WHO only counts deaths that is proven to be directly caused by Chernobyl, while Yablokov includes deaths that are statistically likely to have been caused by Chernobyl.
My conclusion is that 985,000 is plausible, but so is 4000 and there is no way of knowing the actual number. I understand that it's in IAEA's best interest to have the public believe that nuclear power is safe and thus only include deaths confirmed to have been caused by radiation from Chernobyl. Likewise I understand that Yablokov has the opposite agenda and thus includes deaths that can't be directly linked to Chernobyl.
In the end I think it's all about the criteria for attributing deaths to the Chernobyl disaster. If you die from cancer two years before you otherwise would have because of Chernobyl, is your death to be attributed to Chernobyl?