It's not just that. You need a common baseline, a non-moving goalpost to compare performance to. Where and foremost based on which metric do you draw the line for not using maximum details?The hypocrisy of reviewers using ultra settings while at the same time saying that ultra settings are pointless needs to be noted...
Compute throughput? Bandwidth? Price? Relative price (to what?).
Sure, you don't have to answer or even think about this, but reviewers have to. So, what if... reviews for this card would have used medium to high settings all of a sudden? Unfair, because, say, a GTX 1650 was tested with high to ultra? Would you need to test RX 5500 XT 4G in other settings than the 8G version? Or what if you based it on price? 300 Euro (market-aligend pricing after initial batch of "price supported stock") is what was a fairly uppest-midrange card before the price craze.
Can I throw in a car comparison? Would test top speed of a small car on inner-city roads only? 50 km/h (~30 mph) - check, it's fast enough? I don't think you would.
And yes, I know, this would be opening a can of worms.