They did but there was no alternative for the "AI Max", it has much higher requirements for bandwith than traditional CPU RAM. They were investigating the possibility for LPCAMM2 or LPCAMM for that system but it was not feasible to maintain that bandwith while retaining data integrity. For the Desktop it was the right choice. It was designed for local inference applications and there the memory bandwidth is absolutely crucial.

For the new Framework 13 Pro however with an Intel based SOC they have now successfully implemented LPCAMM2.

I think so, because that was during the peak of the ram shortage, and LPCAMM2 was barely produced then.