This should be possible, I think I understand what you are trying to do anyway - a diagram may help if I have misunderstood you.
Is it something like this:
Remote VPN Client → (Peplink A WAN IP) ← (SF VPN Tunnel) → (Peplink B LAN) → (VPN Server)
If so, assuming you don’t have NAT enabled on the SF VPN profile, on Peplink A you would do a NAT port forward towards the LAN IP of the VPN server behind Peplink B.
The exact config for this can vary a little depending on what model balance you have, if you provide some more detail I can maybe make some screenshot examples.
In some respects, that should take care of it, the connection tracking in the Balance should ensure that the return traffic is routed via the path it arrived on, if not an outbound policy on Peplink B would be required that matches the public IP of the VPN client as a destination and sends traffic towards that IP via the SF tunnel.
Depending on the type of VPN server you have behind Peplink B you may also need to look at some of the service forwarding configuration on the Peplinks to stop them trying to terminate the VPN (this mostly matters if it is an IPSEC tunnel relying on NAT-T).