For one thing, average temperature is not much use for a holiday. Places that are weather dependent (tropical islands, beach holidays, ski holidays) generally will have the right temperature at the time of year you're thinking of going, but the actual experience of the trip is less temperature related than precipitation and wind related. A holiday on a tropical island with 5 days of rain is miserable even if it's warm. A skiing holiday with a blizzard isn't much fun, and being on a cloudy beach is only half as much fun as being on a sunny one.
Unless you could mine a probability of sunshine in a reasonable forecast period - and this depends on your travel timeframe - the data is always going to be problematic. That's not a slight on the author, it's a problem inherent in weather prediction. And lets face it, some of the biggest computing power in the world goes into weather prediction.
The cynic in me says you can get just as much information on potential destinations just looking at a globe and picking a hemisphere / latitude based on time of year, and gridding down to a set of locations that way.