Reading [1] again and again, I still don’t understand the following sentence:
With policy routing it is possible to force all traffic to the specific gateway, even if traffic is destined to the host (other that gateway) from the connected networks. This way routing loop will be generated and communications with those hosts will be impossible. To avoid this situation we need to allow usage of default routing table for traffic to connected networks
If you setup for LAN1 to go over Wan1 and Lan2 over Wan2, you might also want for Lan1 to be able to connect to Lan2.
To do that you need to accept traffic without mangling.
By default there’s only one routing table (called “main” in RouterOS) and router looks up destination in there. Any connected subnets appear automatically in main routing table. With policy routing, you get multiple routing tables, and by mangling you tell router which one to use. But these other routing tables are your responsibility, nothing appears there automatically. So if you have e.g. typical multi-WAN setup and you have different routing tables for each ISP and they contain only single route to 0.0.0.0/0 (any address) and you tell router to use this table, everything will be sent to default gateway (= to internet) and router won’t care at all that destination address might be in local subnet, because route to that local subnet doesn’t exist in given routing table.