CPU usage is lower because there are more instances serving your application. The tasks are basically distributed.
For every instance, there is a related vCPU used and a corresponding length of time this instance was used. For example, you have 3 instances that served as follows:
1st instance - 1 hour
2nd instance - 2 hours
3rd instance - 1.5 hour(s)
Assuming all instances are using 2 cores, the calculation would be:
1st instance = 1 hour x 2 cores x $0.0526
2nd instance = 2 hours x 2 cores x $0.0526
3rd instance = 1.5 hour(s) x 2 cores x $0.0526