These stats are showing / guessing only the public nodes capacity but not also the number of txs done through those channels, not counting also the private channels that cannot be counted.

A simple question: how many users of private nodes like Zeus, Phoenix, Blixt, Mutiny, Breez, Electrum you can estimate and let's say each one could have at least 2 channels.

Most of the txs are done from these private nodes, not from public nodes. Public nodes are mostly routing these private txs, but they will never know how many.

Is not LN a beauty because nobody could really measure it / trace it ?

are you confident that LN growth fully explains low transaction costs now?

This is not 100% the cause. Take in consideration also these aspects (and you can see it with mempool glasses):

• NFTs/ordinals are done, they are empty pockets and can no longer sustain that useless traffic. This was more than 50-60% of the traffic.
• coinjoiners are no longer doing endless mixing (IMHO no longer necessary)
• big exchanges were already consolidate their garbage and no longer doing massive traffic with huge fees. Many of them are using bad code not using efficiently the fees and types of BTC addresses
• exchanges now are offering LN withdrawals that help a lot
• batching txs and batching opening channels are now more efficient and we have tools

That's why I am saying that LN helped a lot in reducing the onchain traffic.
But don't worry, the high fees will come back.