Java common ForkJoinPool parallelism

If I run the following code:

System.out.println("Common Pool Parallelism: "
        +ForkJoinPool.getCommonPoolParallelism());

long start = System.currentTimeMillis();

IntStream.range(0, ForkJoinPool.getCommonPoolParallelism() * 2)
    .parallel()
    .forEach(i -> {
        System.out.println(i +" " +(System.currentTimeMillis() - start) +"ms");
        LockSupport.parkNanos(TimeUnit.SECONDS.toNanos(1));
    });

I get (for example) the following output:

Common Pool Parallelism: 3
5 82ms
3 82ms
1 82ms
0 82ms
4 1087ms
2 1087ms

It appears to me as though the common ForkJoinPool is using 4 threads, then they're blocked for a second, then the last two jobs are run.

But as I understand it, the common ForkJoinPool defaults to using Runtime.getRuntime().availableProcessors() - 1 threads, which in my case is 3 so I'd expect to get three jobs printing at ~82ms not 4.

Is it my code or or understanding that's amiss?

ETA: Printing out Thread.currentThread().getId() within the forEach also shows 4 different thread IDs.