We you figure out the amount you will be surprised because the number is really small if you compare it with visa transactions.
I don't think it is realistic to compare bitcoin throughput with the throughput of a major credit card network.
Visa's transactions are stored in a centralized database. By contrast, all bitcoin (on-chain) transactions are broadcast and stored by every bitcoin user with a full node. Granted, LN will aggregate transactions together, but the decentralized nature of LN will not allow scaling to reach "visa" levels.
It's a classic Sztorc "Trilemma" of scalability vs decentralization vs security, and the solution is to make a centralized* Layer 2 that is scalable and secure. Everything else is just a band-aid.
*Note: The owners of this network will be the collective group of merchants who want to settle payments with customers. They would each have a partial ownership proportional to their percentage transaction volume on the layer, just like shares in a company. As thousands of merchants will want to use this kind of network eventually, that means that individual ownership of the total is so small that it can be considered to be decentralized.
In theory, depending on how the LN topology evolves over time, it could eventually have similar throughput as Visa. In order for this to be possible, there would need to be a fairly small number of "banks" that many users open channels with, and each 'bank' is connected to each other (or is also possibly connected to a 'central bank'). This is somewhat similar to how the banking system works today but without fractional reserves (the 'banks' and 'central bank(s)' would need to have a lot of capital, or else they would be unattractive for users to connect to their nodes.
Granted, LN will aggregate transactions together, but the decentralized nature of LN will not allow scaling to reach "visa" levels.
Provided that both have potentially unlimited transactions per second, there's no point to continue arguing which one can serve better in speed.
I don't think the LN topology will ever look like what I described above.
It is clear that LN is much faster than Visa transactions in terms of the merchant receiving their money. I understand that merchants' agreements with payment processors vary, but I speculate that in a best-case scenario, merchants receive money in a number of days, although there is the risk of chargebacks months after the date of the transaction. The question of which setup is superior is not a technical one. Each setup has its own pros and cons, and the costs associated with fraud are borne by different parties in each setup.