Mass Delete for Jira repeatedly runs the configured query (for a number of issues at a time) and then performs the removal. For this reason, if the data matching the query changes, it will affect the execution.
If new issues start to match the query (e.g. they are created when the job is running), they may be removed as well.
If some issues no longer match the query (e.g. they were deleted or updated), they may remain unaffected.
Whether any of the above occurs depends on timing and race conditions. In this case the number of removed issues will be different from the count presented at the moment of job creation. The job status page keeps the counts up to date.