Indeed, but for those which play games with high replay value (roguelikes, boardlike games) is a good measure how "good" is the game before we will get bored from it. I know the game will be idle in some computers, but hey, that will happens in only few machines and in every game - so it won't influence the final result too much.

If I would have that information in any game store page, that would save me a lot of time digging into Steam reviews while doing mental maths.

Take a look at the image. That's the 'magic number' I'm looking for.

Well not really, but with at times only 5-20 people being polled, you get a really unprecise average. And even with more data points, calculating the mean between an honest speedrun and a casual 100% doesn't really tell you much.
I only look at it to see what kind of game length to expect.

Speedrunners probably skew the statistics towards longer playtimes. In figuring out how to beat a level in one minute, they may have spent an hour testing various approaches or honing their skills to pull off difficult manoeuvres, whilst us regular players would just have logged 20 minutes.

Howlongtobeat.com doesn't check clocked times on Steam or any other client that tracks time. It's all self reported.

I assume there isn't an incentive for anyone to capture misleading playtimes onto the site? Other than perhaps a couple of people who get a kick out of that sort of thing when they've run out of other "fun" things to do.