> First - that's a big if
That ain't an "if" at all. The actual identifying information on, say, a vehicle title or land deed takes up maybe a couple hundred bytes at most (if not far less), whereas most NFT-capable blockchains allow transaction metadata on the scale of multiple kilobytes.
> Second - do you realise that title contains private information which will be forever exposed on-chain?
Nearly (if not entirely) all of the actually-private data (namely: the owner's PII) is redundant under a title-as-NFT system, since ownership is asserted by wallet/token custody rather than being baked into the document.
Besides, encryption is a thing that exists.
> Third - ok, even you admit that not everything will fit on-chain.
I admit no such thing. What I actually said is that you can break up large data into smaller chunks such that it can fit on-chain even if it somehow blows out the chain's transaction metadata ceiling.
Regardless:
> Then who will store the rest of the data?
Have you heard of IPFS?
> Fourth - in the hypothetical case when we have titles and NFTs linked to them, there are several interesting questions - what happens when owner of the NFT forgets his private keys? What happen if the owner dies? What happens when the court judges declare that the ownership is incorrect? How are marriages handled, where today multiple people can own the same property?
All of these things can be (and, in the case of lost keys, have been) addressed with smart contracts; the blockchain can programmatically define conditions under which the NFT can be transferred, e.g. with consent from multiple parties. Such programs can also represent multi-person ownership; in a world where blockchain computation can handle the governance of large organizations, something like a marriage or an estate is a readily solvable problem.