I suspect poor forecasts are due to KPIs set by people who don't understand how the forecasts are used.
If you forecast prob 30 fog, and it only happens 25% of the time, it counts against you in your accuracy KPIs.
Whereas if you forecast CAVOK, and amend the forecast for fog 15 minutes before it rolls in, it counts as an accurate forecast.
This was the impression I got talking to a forecaster at one of the CASA information nights. They did not seem to understand why a late notice amended forecast could pose a problem.