Counting is harder than it sounds and the difficulty is not squeamishness about a number. It is that the number does not exist in one place.
Witch trials were held in an enormous variety of courts across three centuries and dozens of jurisdictions: episcopal courts, secular criminal courts, seigneurial courts, town councils, and in some places bodies with no permanent existence at all. Some of those courts kept careful registers which survive. Some kept registers which were later destroyed by fire, war or indifference. Some kept nothing. A count therefore proceeds territory by territory: a historian works through what survives for one region, counts trials and verdicts, and produces a local figure with a stated margin. Aggregating those local figures into a European total requires extrapolating across the regions where nothing survives, and that extrapolation is where the disagreement lives.
There are further complications that are easy to overlook. Not everyone tried was executed, and acquittal rates varied enormously, so a count of trials is not a count of deaths. People died in custody before verdict and are recorded unevenly or not at all. Some jurisdictions executed for related offences under other names, which is how Anna Göldi was condemned as a poisoner in 1782.
The consequence is that responsible scholarship gives a range, states what it rests on, and does not offer a single figure. So does the atlas.