The original problem, as stated by me, was:

"You are generating k random 64 bit integers. Each time you generate the integer, you have to insert it in a sorted collection.

(...)

The question: in your estimation, both algorithms being implemented optimally, what magnitude k needs to be for the linked list to start being more efficient than array list."

The cost is not evaluated for any individual insertion, only for the total cost of inserting k integers.

For some of the integers the cost of inserting to linked list will be higher, and for some the cost of inserting to array will be higher. It does not matter. We do not care. We just care about what is the cost of inserting k integers thus randomly generated a) into sorted linked list and b) into sorted array list.