In the terminal phase, it's a straight shot with proximity detonation of the 2000 kilo warhead. But at a mile in less than six seconds, a large ship has very little time to react once the things lite off...less than a minute for 10 miles.
Several variants. Some are seeded like limpit mines and sit in hibernation mode at 1000ft waiting. Others cruise to the target acquisition area at 50kts and then go into acquisition mode. They target a vessel and then lite off when they are close enough to assure a hit. Some launched by aircraft, etc. Later, a variant is developed that does a pop-up manuever where it actually lifts out of the water and attacks accross the last couple of miles like a missile.
They use digital terrain maps of the ocean floor and have sophisticated sonar and acoustic target acquisition capabilities.
All fictional of course.
EXTREMELY fictional. There's no way in hell one of these things will actually go 450 miles UNDERWATER and hit a target--maritime targets have this annoying habit of MOVING.
Read up on Outlaw Shark, and the oopsies that happened with that program, to get an idea of (1) the isssues involved in OTH-T, and (2) why we canned the TASM. And the TASM could actually receive target location updates while in flight. Your supercavitation weapon can't.
So, you're stuck with either (a) shooting at targets that can't move at all, (b) shooting hundreds of these things for one hit (which would bankrupt anyone except the United States, and we'd have a hard time maintaining that rate of expenditure), or (c) get used to your weapons locking onto anything that's even remotely close to the intended signature (i.e., your own Super Panamax container ships full of rubber dog poop and whoopie cushions, oil and LNG tankers inbound to Shanghai, et cetera).
Half of your problem is that you used Harpoon for modeling: that game system's EMCON modeling is extremely poor. Communications are never affected by EMCON, and are never counterdetected--you can have a submarine ghosting an enemy surface group via periscope, transmitting detailed OTH-T data without the enemy knowing they're there unless they get a hit on the periscope via radar. It grossly oversimplifies the entire complexion of scouting and antiscouting. Reality does not work like that. It also heavily abstracts the midcourse guidance issue in a manner that makes hits far more likely than they would be in reality.