The problem of this kind of benchmarks is that they are misleading.

In a real case scenario you probably are going to query a DB or do something, which will make the extra time express requires irrelevant.

It is like comparing [].forEach vs normal for loop. If you do something inside the for probably the time for the function call of the forEach will be become irrelevant.

One may think that using express the API call will be 50% slower... which won't probably be the case.