Is the FET LTspice model good to estimate loss?

I try to simulate a buck converter to get the loss. It is under ZVS or partial ZVS. The result i got is a lot less than test. I am wondering if the model is accurate enough to get the conduction and switching loss.
Under ZVS or partial ZVS, is it still good to use EPC online loss calculation for buck converter tool?

yes, LTSpice models are quite accurate for conduction and switching losses, especially to help in comparing and selecting devices. For ZVS, this method will be much better than using the online buck calculator which assumes a hard switching buck.
For actually matching with real world data, much more work will be needed because the rest of the circuit elements (like inductor, capacitors, gate driver) as well as parasitic elements (like layout parasitic inductances and resistances) have a large influence on the overall losses. Models for these elements are often difficult to find and/or generate, and this significantly limits the accuracy of the simulation.
An example of these effects can be found in the book by Lidow, Alex; de Rooij, Michael; Strydom, Johan; Reusch, David; Glaser, John. GaN Transistors for Efficient Power Conversion (p. 90). Wiley. Kindle Edition.

In this case a good match was found but it required to add all the elements in the real circuit.

Bought the book, reading the book. Great detail! EPC is doing great job education us with all the technical details through books, app notes, videos. Really appreciated.