Ok, let's try explaining this a different way.
The likelihood of a human getting injured by a dog is basically down to 2 things.
First is how likely that dog is to attack in the first place, how aggressive they typically are, how often other dogs of the same breed attack humans (not other dogs. Humans).
The second is how much damage the dog can inflict when it does attack. This is what you are utterly fixated on (albeit with a very bizarre logic).
When you look at both things together,you could create a rough 'danger score'
So to work out the likelihood of a dog attacking in the first place, you could maybe look at how many attacks there have been in previous years, taking into account how prevelant that breed is.
Sausage dogs and chihuahuas don't feature in the top 10 most popular/prevalent breeds in the UK, but they do feature in the top 10 for bites. So you can reasonably say that those dogs are aggressive. The likelihood of being bitten by them is well above average.
Staffies are either the most popular or 2nd most popular breed in the UK, depending on how you measure it. Yet they don't feature in the top 10 for bites. So you could reasonably assume that you have a very low chance of being bitten by one.
Where your logic seems to depart from everyone else's is how much weight you give to the likelihood and how much to the potential damage, when calculating the overall risk. I personally find it easier to live with a very very low likelihood of something going wrong, even if the potential 'wrong' is worse.
Put another way, you and your children have to make a journey. You have the choice of 2 cars to get you there.
Car 1 is being driven by a drunk driver in a brand new car with loads of airbags and safety features.
Option 2 is a very experienced driver in an older car with no airbags.
Which car do you get in to?
Your dog example is car 1.