I think the confusion is due to what the portal means by “common essential” and how the “common_essential.csv” file is defined.
On the portal “common essential” is defined by looking at the CRISPR data:
So, if a portal flags a gene as “common essential” it means that it looks essential in this particular dataset.
However, the “common_essential.csv” file is defined as:
or, in other words, these are genes that Biomen and Hart have published as “common essential”. (We use these as our reference set used to scale gene effect, so we want these to be independently derived)