Well, this goes back to 16-bit addressing (as revealed by your use of near).
In 16 bits, the size of an int (being 16 bits) is 2 bytes.
You increment it twice. But the compiler knows it's an int pointer, so it increments it by 2 each time.
So you're adding 4 to a value already holding 0xFFFF. Adding 1 to 0xFFFF in 16 bits rolls over to zero. What do you figure you have left?