Also your logic is flawed: If only big transactions are viable on L1, then the transaction fees must be much higher than now. TX fees as of now are about 5% to 10% of miner revenue. After 2 halvings more (i.e. when you argued in another thread that Bitcoin mining will collapse), tx fees even on the current level would be thus around 20-40% of miner revenue. In your "only big txes are possible" scenario, I guess we'll have fees at least 3x or 4x higher than now, so the fees would actually be similar to the block rewards after the ~2028 halving or even similar to now.
But I think L2s are indeed a better idea. Why should every full node have to validate, relay and store each coffee somebody has drunk ... or the activities of a weather app storing all temperatures of a lot of stations, like it was the case on BSV?
And ... fees would actually be much lower in a big block scenario. Look at the fees at BCH and BSV. Why is not everybody using these chains? LTC, which has a relatively small-block model (although a slightly higher capacity than BTC) is much more popular for transactions than the big-blocker forks.
Or any other receipt network.
But Lightning is deeply rooted on L1, much more than other ideas like sidechains. For small amounts it works fine. The theoretical attacks that have been discovered (and yes, they are hard to fix, in the case of the Replacement cycle attack) are only viable if they could be used to steal big amounts.