How good were they for hurricane forcast before 2003?  How do they forcast hurricanes? If they forcast the number of hurricanes based on the data about climate patterns and incidence of hurricanes in the last fifty years, this won't work well because of rapid climate changes in recent years. The news said that this winter season we experienced the warmest winter in Canada on record (http://news.yahoo.com/s/ap/20060314/ap_on_sc/canada_warmest_winter_3;_ylt=AhgHyFUFqXxMHBNB57wztw5rAl MA;_ylu=X3oDMTBiMW04NW9mBHNlYwMlJVRPUCUl).  Then we technically entered an uncharted territory of climate in terms of temperature in North America this year. Then how can they forcast weather six months later without having any data relevant with the current climate conditions. The only way to do this is to set up some sorts of models and to extrapolate the past data to the current situation. But I doubt that we have good models for weather forcast.  
Interestingly, the seasons in the 2000-2003 timeframe where much milder and don't to the casual eye look like much evidence of anomaly. (You can check out the archives). So either 2004-2005 have a strong component of unlucky fluke to them, or we are crossing some kind of non-linear threshold...

One possible factor is the state of the loop current in the GoM.

Thank you for your reply. I think that the climate is the result of combination of chaotic random factors and predictable factors which follow laws of physics. As many data indicate, the temperature has increased globally in the last one hundred years. Although this is merely a trend, it is very unlikely that the temperature next year would drop down to the level of temperature one hundred years ago. This is predictable because the temperature must obey the laws of energy. The size of polar ice is pretty much dependent on the temperature in the polar region. Therefore the size of polar ice is one of good indicators of global warming. On the contrary, the number and strength of hurricanes are dependent on many factors including both chaotic random factors and predictable factors. When the sea temperature is higher, there are more chances for hurricanes to become stronger. This is predictable because hurricanes are getting energy from heat energy retained in the sea. However, it is unpredictable how, when, and where hurricanes are formed because this belongs to the realm determined by chaotic factors. There might be some patterns in the relationship between hurricane formation and other climatologic phenomenon. But I think that these patterns are only temporary at their best.
But chaos can be predicted if we have a big enough chaos theory predicting program.  But this is not true chaos, this is a big set of variables that we have not yet wrapped our hands around to see the outcomes of the things that are going on.  Cause and Effect are happening on a vast scale, but they are happening in a Logical manner even in the oceans.  So technically speaking its not Chaos, Just not yet understood by us humans.

Maybe I have just written to much science fiction and Read to many papers on Chaos and how to make the complex simple.
I almost feel as if I know where this is all heading but can't quite put my finger on it.

It is very hard to conclude what the boundaries of outcome are for this chaotic system, and how can we conclude if the chaos inside these boundaries is predicatble or not? It might be like a very simple multiple hinged pendulum problem, we can predict how the swings will peter out on average but not in wich direction it will swing a number of swings into the future since imperfections give changes that gets larger for each swing. The need to know everything about the starting conditions is sooner or later limited by our ability to measure.
I have no doubt they are underestimating -- however I don't think you can look at only two years and call the results in to question. Also: almost all long term weather forecasts are bull