Wait a minute... Before comparing one year with another, using data such as those in the post from the hawk, we need to standardise the numbers of accidents for activity level. That is to say, we need to compute the accident rate, not just the raw number of accidents.
There is a variety of ways of defining accident rate. It could be number of accidents per cycle, or per passenger kilometer flown, for example. It's a little worrying that, in their announcement, B.A.A.A./A.C.R.O does not seem to have done this, but maybe the rates are given somewhere else in their data base.
Of course, our intuition tells us that the number of cycles and the number of passenger kilometers were both higher in 2007 than in 2006, and so if the raw accident data show a decrease then so too should the accident rates. But the numbers that would go into the denominator (that is, the number of cycles, or the number of passenger kilometers, etc) probably haven't been increasing for each year over the last decade (consider the post 9/11 effect), and so we need to be cautious in interpreting long-term trends.