There are various reasons why people come up with different numbers:

The first question is what are you actually counting. The most usual approach seems to be soldiers killed in battle plus civilians killed either by direct impact of the war or indirectly by famine and disease caused by the war. Victims of the holocaust and other nazi mass murders are included, as are victims of German and Japanese war crimes. The timeframe is usually from the first act of war in Europe, Hitler's attack on Poland, to the Japanese surrender. There's a few points right there which may lead to different estimates, the most variance is probably in estimating which civilian casualties are war related. Civilian casualties of strategic bombings or people directly killed by the advancing Wehrmacht (or later the Red Army) are pretty obviously war casualties, but when it comes to disease and famine it's not always that easy to judge whether a death is war related or not. Was a death by disease just basically bad luck or was the disease only deadly because of malnourishment directly caused by the war? Do you count soldiers wounded in the war who die from their wounds after the war has ended? Do you include victims in the war between Japan and China before the war in Europe started? What percentage of missing people do you presume to have actually died?

Then there is the fact that a lot of the documentation and paperwork you would need for exact numbers was destroyed in the war, or possibly never existed in the first place. And during the war, gouvernments (especially in Germany and the Soviet Union) would not have been keen on making exact casualtiy numbers public (or, quite frankly, casualty numbers that were anywhere close to reality)

So it's not really surprising that the numbers diverge a lot. There's cases where you can pin down things pretty closely, like the number of US soldiers killed in action (you will still need to make estimated considering those missing in action, but you will get close to the actual number), and then there is cases like the Yellow River flood in 1938 when the Chinese Nationalist government intentionally destroyed the dykes to stop the Japanese advance, where you can only really say for certain that hundreds of thousands were killed directly (estimates vary fromm 400.000 to 900.000), and then you would still have to decide how many deaths in the aftermath of the flood you want to attribute to the war...

Usually the variations come from who they're counting. The 20 million range is direct military casualties. i.e. just the people that were in armies. The 50 million range includes all the soldiers plus the civilians that were killed by military operations. i.e. civilians killed in fire bombing of cities. The highest range, the 70 million range, includes everyone, the soldiers the civilians killed in battles, and everyone that was killed due to famines and disease outbreaks that were caused by the war.

When I learned about WWII, figure stood at ~40 million, roughly 20 million in Asia and 20 in Europe (most dead being from China and the Soviet Union).

Estimating these things can be tricky. The perpetrators aren't keeping exact records of how many they kill. In places like China, which was racked by war before the Japanese invasion, the government might not be aware of the exact number of people living in a certain area before it was destroyed by the enemy.

It's also a matter of defining what counts as killed due to a war. Sometimes historians look at "excess deaths" during a period as causalities of war and sometimes they look at a drop average life expectancy as a measure. WWII devastated the economy of the Soviet Union, so maybe it is fair to measure people who died earlier due to not getting health care because of this as casualties of war.

Its probably a couple of different factors, though I'm by no means an expert. 1. Exactly when do you start WW2? 1939 and the invasion of Poland? 1931 and the invasion of Manchuria? Somewhere in the middle? Adding 8 years and a hot war will change the numbers substantially. 2. China was also in the middle of a civil war; how many of those count? 3. The numbers are always squirrelly in wartime, especially civilian deaths in areas where the records were also destroyed. And there have been strong incentives to "adjust" casualty figures for political ends. Stalin and his successors would inflate casualty counts to show that the West were freeloading off of Soviet casualties or blame deaths they caused on the Nazis, China and Japan try to "out-victim" each other wrt to deaths in Nanjing and Hiroshima+Nagasaki, etc.