I think the question I am asking is: is there an alternative approach to what Ethereum is solving with the Oracle where external data is fed into the blockchain to be reused inside the chain itself afterwards. This data comes from a central server, that holds alone all the conditions to this data but is trusted by the network.
No. There are some misconceptions here.
First, there is no working Oracle yet. For now, there is no way to insert outside data to the blockchain yet.
The way blockchain is designed and its validation works, you cannot feed outside data to it now.
However, there are some projects working in a solution to this problem. Some of them are centralized like you mentioned (which defeats the whole purpose of bitcoin and ethereum) and there is one that is not centralized, which became famous recently: chainlink.
Chainlink is not operational and it is decentralized. Anyone can run a node and collect any data from real world and send to other nodes for validation. It will also validate data.
But neither chainlink or those other Centralized projects are working yet on ethereum. So you cannot use them
I doubt vitality and ethereum community will ever accept a centralized Oracle.
After every hour I shoot my data of apples on the blockchain through a (trusted centralized?) transformation method somewhere. How can the other nodes verify that this centralized transformation was executed correctly. There is no alternative way to predefining a trusted source to do the calculation and blindly trust the result I suppose?
There is no trusted source in blockchain. Forget this idea. Blockchain is trustless. You will need a decentralized Oracle to feed your data, or insert it all in a smartcontract and replicate it in all nodes (if. Possible).
Maybe you need another technology. What kind of data do you want to feed?