Simple, because when L1 does not work as intended, people created L2. When L2 still has problems, people will create L3. And when L3 will have another issues, people will create L4. If L1 would be good enough, people would never create L2, because it would never be needed (as you can see on many altcoins where you don't have L2 or where you will never see L2 by design).

L1 does not work as intended, because if you use BTC extensively and make many transactions, then they have to be stored forever. There is block size limit, so mempool is sometimes congested when there are too many transactions. For that reason L2 is needed. On BCH you have no L2, because they decided to do everything on L1 and they promise to never introduce L2.

L2 does not work as intended, because when L1 is congested, then you cannot settle your L1 transactions. You cannot open and close channels without L1. You cannot send coins to someone who never used BTC when L1 is congested. You cannot close your L2 channel for one satoshi per byte when there are 100 MB other transactions waiting in mempool.

L1 does everything on-chain. L2 does channel opening and closing on-chain and everything else off-chain. I guess L3 would batch matching channel opening and closing on-chain transactions into single on-chain transaction. From L1 it would look like channels will be closed before opening transaction will settle on-chain. I guess L4 would push some kind of 256-bit commitment to the nearest block and everything would happen off-chain. From L1 it would look like some coins are still part of the chain, but no L1 miner knows who owns what in L4 and how many channels are there, the only visible thing would be that there is X BTC locked in the whole L4 layer. But of course L3 and L4 design is just guessing and I am not 100% sure how it will work in practice.