The assumption behind the cohesion metrics is that methods are related if they work on the same class-level variables. Methods are unrelated if they work on different variables altogether. In a cohesive class, methods work with the same set of variables. In a non-cohesive class, there are some methods that work on different data.
In the metric, number of pairs of methods that share an access to instance variables, fields, is subtracted from the pairs that do not.
But what about methods that simply perform calculations and return a value? I have a lot of them but as they do not share any instance variables, it is recommended to separate them, which does not make sense to me.
But what about methods that simply perform calculations and return a value?
Those kind of methods you are referring to are called utility (aka helper) methods and usually are considered code smell from OO design perspective. The following article elaborates on the topic of utility methods and provides an illustrative example with a refactored alternative: http://www.yegor256.com/2014/05/05/oop-alternative-to-utility-classes.html
Probably, your static code analysis tool doesn't recognize utility classes (class which contains utility methods only) and gives you the same hint it would give for a misplaced method in a class. Moving each method of a utility class to a separate class wouldn't make much sense, assuming that you've already grouped your methods based on some criteria.
My recommendation for you is to revisit your design and try to eliminate the use of utility methods. If for some reason this is not feasible, then keep your original grouping of utility methods.