"A game that could look almost as beautiful as CryEngine/Unigine capabilities?"
No not really, at least not if you want the engine to be exactly as these two engines as many, many people have worked to get to the point as it is now.
If you only want the graphic then perhaps, but if the physics, sound, AI etc. etc. has to be a part of your engine too then no. The problem is also how much time you wish to spend on it. Grab a MIT engine and make it your own by tweak or just use it for what you like
I understand, but I do not want to make a game engine, the game itself will be the game engine. A game engine is more than just graphics, physics or the low-level stuff, right? They need to be usable in a lot of games, they need to provide tools for other people to use, easy workflow and so on. That is a hard thing to do that takes time and knowledge and many years to get it right. They also need to provide innovation to sell their product. Cross-platform too.
Of course, it takes a few years to learn and get the experience you need to build a 3D game without an existing game engine, but isn't that worth it? Is not like all the knowledge accumulated is for nothing. Also, is not like everything will be build from scratch. For example a free physics engine could be used, like Bullet or PhysX, right? Maybe other components also...
Now, I know it's hard and takes time and the advice is not do it and I understand why. But, is that written in stone? I mean, there is really no reason not to do a 3D game without an existing game engine? Is the knowledge present for everyone willing to learn or you need to be really smart?