When I train sprints, load estimation based on HR is very wrong. At other workouts it is OK.
Is this correct?
Wow your sprints are way above the rest of your power curve which is breaking the model a bit. I can do a similar workout to that one but my FTP is 25% higher than yours. You can see on the power page that you are way over the “typical” curve.
If you do a sprints with only HR the training load is going to be wrong. But normally sprints are done with power, HR is ok for endurance and longer intervals and the model will work well for that. So I think it’s ok.
I know that I have quite good power for short sprints, the question is what to do to improve also on long distances 
EDIT:
morton 3 parameter model:
eFTP 161w W’ 43720 J
Defaults:
eFTP 240w W’ 20123 J
Which one is better for me?
And somehow sprints are great for me
The Morton 3 parameter model only work well if you have the right inputs. I think its 5s, 3m and 12m all max efforts. So it is very likely underestimating your FTP. So you should go with eFTP and do some longer efforts 
An FTP of 161 and peak power close to 1000 Watts, with that kind of high intensity energy, is near impossible. I would think 240 / 20123 makes more sense…