There are many factors involved…
MNDP is encapsulated in a UDP packet with an IP destination of 255.255.255.255 (this network), sent with an ethernet broadcast destination address so is propagated to everything within a layer 2 network.
CDP is encapsulated as LLC/SNAP with an OUI of 0x00000C and protocol ID of 0x2000, sent with an ethernet multicast destination address of 01:00:0C:CC:CC:CC. Cisco devices do not forward this so only see CDP packets from directly connected devices, other manufacturers typically do not treat the destination address as anything special so the CDP packets are propagated to everything within a layer 2 network.
LLDP operates at the interface level. Any packets with a destination address of 01:80:C2:00:00:0x received by an 802.1D-compliant bridge are never forwarded to other interfaces, only the device CPU, so you only see LLDP packets from directly connected devices.
Mikrotik bridges appear to handle LLDP correctly unless protocol-mode is set to none which disables compliant operation.
MNDP and CDP packets which are received directly by the CPU from the bridge identify the correct interface, I believe this is due the ingress interface being noted internally for bridged.
The main issue appears to be where tagged packets from the bridge-to-CPU interface pass through an /interface vlan to remove the tag and present the traffic to the CPU, I suspect the internal noting of the ingress interface is lost causing neighbours to use the vlan name as the real interface is no longer available.