I think that it is an accurate way to test.

Appreciate your thoughts, but the reason that statement is wrong is because (as his title states) he is testing BTC and BCC transaction times. This test is not accurate because he is using a control group as a consistent (the exchange) that he has no control over.

It's the same as saying "I'm going to test which airplane is better by seeing which one makes it to my city first. Ready, Go!" Obviously whichever company gets the plane off the ground and into the air first is going to have a headstart on getting their model of plane to you first.

So really he's testing the company not the product. If his title were "How do Bitcoin and Bitcoin Cash transactions compare when sent from ABC Exchange?" Then it would be a more accurate test. And even that test would be worth about nothing because you'd have to collect a ton of different data points before a true average set in.

In general we are only interested in the user end experience

If that's the case, with this test, your subject is the Exchange not the coins and as such no opinion at all should be drawn about said coins. That's not what's being tested here.