The tunnel interface IP is used for the traffic originating from fortigate itself to travel on the tunnel.
Without that, this traffic would use the exit interface IP, which would be the WAN interface IP and it would not be able to go on the tunnel
The reason we do this in SD WAN is because fortigate is sending traffic for SLA or ping for checking if the link is still up
To allow all this traffic to go across the tunnel, we have to define tunnel interface IP and add it to the phase 2 selectors.
So now when Fortigate sends all this traffic on tunnel, it would use the tunnel interface IP.
Here are some articles which you can refer for more information: