First link is about use case where ISP delivers internet via PPPoE … and PPPoE can be terminated either on CPE or customer’s own router. This works because PPPoE is entirely different protocol (a point-to-point tunnel) and IP address, associated with local tunnel endpoint can move according to tunnel termination point. So when CPE is configured as bridge, it mostly means that CPE itself doesn’tt start PPPoE session/tunnel, instead simply passes PPPoE packets between its WAN interface and its LAN interface. According to your vague description this case doesn’t seem to apply to your setup. And, if you’re concerned about latencies, PPPoE adds some latency just the same as any other tunneling would do (IPsec, Wireguard, …).
The second link is (in theory) usable in your case as it seems that your WISP uses all-IP network. However, if you disable NAT on your router, WISP will have to configure routing for IP address space of your LAN all the way between your CPE and their NAT router. Additional potential problem is if your LAN address space overlaps with another subnet address space (either of another customer like you or even WISP’s own subnet) which makes it impossible for routing to decide which LAN should be target of packet with dst-adddress set to one of “problematic” addresses.
I understood the second approach. But in the first approach, can’t I handle the PPPoE with my mikrotik router if they put CPE in bridge mode? Why doesn’t it apply to my case?