I agree that Construct is a very capable tool that CAN produce hit games like a Hyperlight Drifter, or a Hotline Miami, or a Nuclear Throne, or a Celeste, or a Dead Cells, or a Skul: The Hero Slayer. There's no reason from a technical standpoint that Construct can't create those games. The issue is that those game designers are choosing either GameMaker, Unity, or to create their own engine.
The highest grossing game I have seen created in Construct is Hypnospace Outlaw, which has generated some pretty good cash but is no where near the success of the games I named above.
IMHO, I believe the reason that the best indie game designers are choosing other game engines is because of how Scirra markets their product. Scirra uses a more quirky style with weird and cute cartoon characters on their landing page. It gives it a more childish feel, like it's not meant for "serious professionals". They also don't highlight games made with Construct very well. Literally the only game they showcase is "The Next Penelope", WHICH CAME OUT 4 YEARS AGO AND DIDN'T EVEN SELL THAT WELL. They need to keep updating their showcase with good looking fresh games, but they do not.
Now if you look at GameMaker's landing page the presentation of it is much more clean and professional. Nothing cute or goofy to make it look childish. They also showcase good looking games that have sold lots of copies, like Hyperlight Drifter, which communicates "these people made tons of cash with our game engine and you can too". They even showcase games like Flynn: Son of Crimson and Hancho, WHICH AREN'T EVEN RELEASED YET BUT THEY LOOK AWESOME.
Another thing I think has plagued Construct from becoming popular was the inability to use an actual programming language for so long. GameMaker and Unity have always had GML and C# scripting respectively in addition to visual/drag&drop functionality. Good experienced programmers want to use a programming language they are already skilled in rather than learn some unique proprietary visual event system where those skills won't transfer to anything else. Now I realize Scirra has already changed that by introducing real Javascript programming, which is a huge step in the right direction. But I think not having that for so many years has already solidified the perception in people's minds.
I believe Construct CAN be a popular choice for experienced game designers who make hit games in the future but they need to make some changes before that will happen. First they need to change their branding and how they market their product into something more clean and professional. Then they need to communicate very widely and heavily their new Javascript programming feature to appeal to experienced programmers. Lastly they need to keep their showcase of games updated and fresh with games that not only look good but actually sell a lot of units. If no games exist that have sold a good amount of units they need to help and encourage indie devs to do so through education and support. They need some brand ambassadors that will make training courses on more advanced topics and will demonstrate how to make games similar to those that have been hits in the past.
Keep in mind that these are just my opinions. But I have used Construct 2/3 for about 8 years and have always pondered this exact same question.