Technically, I could generate a key pair and have a totally legit address and it not be part of the blockchain (offline wallet). Unless you're saying every address possible is already part of the blockchain.
Yes. But in the very beginning of this topic the words were "... statistical analysis of the Bitcoin public keys on the blockchain ..."
Tell me what you want to achieve, if you perform analysis on the data you generated yourself?

Just a proof of concept of finding an address generation flaw with a few hundred thousand of addresses between each "flawed" generated address.
I think we've gotten pretty off-topic here, but I've learnt a lot about the blockchain and more information on bitcoin overall, so thanks a ton!