Hello Attila
The goal here is to determine the L2 address from the L3 address. We have the IPv6 unicast address of the destination, but we don’t have the L2 address of the destination. There is no mechanism in place that can directly determine the appropriate L2 address from the known IPv6 address. Therefore, by definition, we cannot use the IPv6 unicast address.
However, we want to inform the destination that we need its MAC address. The NDP gives us this ability through the solicited node multicast address. This process does have a mechanism in place that can be used to directly determine the multicast L2 address that corresponds with the intended destination. By using this multicast IPv6 address, through which an appropriate L2 address can be derived directly to contact the intended destination, the MAC address of the destination is requested and obtained.
So you see, it is a matter of what mechanisms are in place to enable the direct derivation of the L2 address. The IPv6 unicast address by definition (and by design) cannot be used for this purpose, but the solicited node multicast address can be used. Does that make sense?
I hope this has been helpful!
Laz