Ok it is not feasible in a way of knowing 100% the exact connections. But in places like Australia, New Zealand, Japan ect where the density of nodes is small (nodes are isolated from the major node volume in Europe and USA), can node connections be estimated with a good approximation?
Are any secret nodes that can not be detected on the btc node map?
Why do you assume those nodes are isolated? They are assuredly not— the software makes some effort to form topology-diverse connectivity and beyond that the network is randomly wired.
The overwhelming majority of nodes are 'secret'. Nodes which do not accept incoming connections never have their addresses visible except to the nodes they directly connect. Non-listening still function as complete nodes relaying transactions and blocks even though they aren't directly observable.
There are also a fair number of nodes which connect over Tor, and a number of nodes which are only inbound reachable as tor hidden services.
The bitcoin distributed algorithm is also not married to the peer to peer protocol, it's perfectly reasonable to link nodes via alternative means— e.g. bulk transferring blockchains (via https, scp, torrent, etc) is fairly widely done. People could even ship around transaction steganographically embedded in cat pictures. (and I've seen transactions being transported via IRC and pastebin).
You could certainly make measurements of the public stuff and infer some likely things about the hidden connectivity by things like transaction propagation times... but you can't reliably measure much of it.