Quartus result not making sense, multicycle resulted in large violations?

Did you add a multicycle -hold 1 to the same path? Without that, you're telling Quartus the datapath must be greater than 10ns. With the multicycle -hold 1 you're saying it can be anywhere between 0ns and 20ns, which I assume is what you want.

(Because of variation between max and min timing models, it is almost impossible to have a path squeeze between 10ns and 20ns, so having just a multicycle -setup 2 without the multicycle -hold 1 is often impossible).