First of all, we're talking about plasma and not whole blood - the difference is important.
> An adequate and reliable supply of safe blood can be assured by a stable base of regular, voluntary, unpaid blood donors.
An adequate and reliable supply of plasma cannot be assured by a stable base of unpaid plasma donors - countries that pay for donations run a surplus and end up supplying it to countries that don't, because the latter don't produce enough.
> if the screenings were totally reliable why would we bother asking for risky behaviours ?
Because it saves time (and money) to be able to rule out people who are ineligible for blood donation before they ever get their blood drawn. You might think, "how could a person who has had Hepatitis A not know that they are ineligible to give blood?", but the answer is that this happens all the time.
Screening questions are there as a cost-saving measure; they're not intended to catch anything that blood testing wouldn't, and they're already arguably overly conservative in the screening stage as it is.
> What about the window between an infection by the HIV and the production of antibodies ?
>
(A) For plasma, you don't test for antibodies - you test for viral load.
(B) The window is incredibly short. The chance of someone donating plasma within a few days of being infected with HIV and having a viral load high enough to transmit HIV but also not test positive is minuscule. At that point, there are far bigger risks to worry about, like the many things that you can't test for (and which the patient may not know they have). Risk management isn't about eliminating risk; it's about systematically and contextually evaluating the appropriate levels of risk.