Disclaimer: I run a Brussels-based startup accelerator which, as part of its program, assists startups in these applications (just like those consultancy companies you mention).
And I completely, entirely agree with you. The process is very, very annoying, long, bureaucratic, uncertain and frustrating.
It's quite transparent actually unlike GP says, but the main problem is that the scale of the complexity involved in those applications means you are going to be paying consultants to help you as part of the process. This means there's a bunch of money flowing towards pure paperwork, and often those consultants also take a success fee which reduces the total amount paid out to the project.
In other words, if you have a 1MM EUR grant, on average 100-150k will be spent ... on the application process across the consortium. And this is just for the winners; given that there's a lot of entities applying to these, I wouldn't be surprised if the "application spend" reaches ~30% of the total. And while not all of it comes out of the grant budget, it's spend triggered by the grant.
Cascade funding (https://en.wikipedia.org/wiki/Cascade_Funding) is .. kind of better? Because it allows smaller projects to exist and apply on their own without as much hassle. The smaller calls are a LOT more competitive however so you don't escape this added work created by the grant application: Many many person-months end up being wasted on the process. But the thing with cascade funding is that every layer of funding that gets cascaded takes a % in operations, and does create additional "work" in the system: It's just spread over more entities and projects. So while it succeeds at "moving more money", a lot of the value-creation is pure busywork.
TBF I don't know a single well-meaning EU civil servant that doesn't think the system needs to be reformed. But into what? How else do you distribute these huge budgets, in open, transparent and democratic way to so many people without creating these huge and complex systems, all while maintaining good checks against fraud? It's easy to say "we can do better" especially here on HN but solving this is magnitudes harder than the types of scale problems Google & Meta face, and it's cross-disciplinary not just engineering. The EU has funded really fantastic and inspiring things, we shouldn't dismiss the whole thing outright.