Because 1) Boinc projects are not suitable for ProofOfWork (hard to find, easy to validate); 2) Boinc projects were/are centralized; 3) The Bitcoin creators 4 years ago did not anticipate and are not responsible for today's "insane waste of computing power"
I don't think 1 and 2 are always true -- for example, take protein folding. It's hard to calculate a protein's ground state, but easy to validate against NMR or crystal structures. It also doesn't need to be centralized; anyone can plug into the Protein Data Bank and get a full list of experimentally solved structures.
The Bitcoin creators did not anticipate that Bitcoins would blow up the way they have, but clearly, they knew that they were wasting some computing power, and that the amount would grow if/when Bitcoins became popular.