I think in your equation you are probably missing technological adjustments (so that same hashpower => less energy required).
the tx fee is not exactly connected to the exchange price, is more something which depends on the congestion of the network.

Layer 2 solutions, like lightning, can mitigate the weight of tx on the layer 1, so that, eventually, more and more tx will happen on a layer 2. I suppose it is going to be "safe" to move up to 10k$ on lightning.
(maybe I am wrong on this, it is just a quick mind evaluation)

For tx of higher values it will probably still make sense to do it on layer 1.

In the case it will not be, as you suppose, worth to send smaller amounts, probably different types of services will pop-up (escrows, insurances, or whatever our creativities will come up with)!

Your point is not wrong, but it's probably missing some "organic" factors.

And maybe I am being too naive!

Who knows!