I've got 67 just using random placements with heuristics, programmed in Python. Apparently they've got 73 (which is 289/4 rounded up), and that's why they think this might be possible.
1 1 . . . . . 1 1 . . . . . 1 . .
. 1 1 1 1 . . . . . . . . . . . .
1 . 1 . . 1 . . . . . 1 . . . . 1
1 . . 1 . . 1 . . . 1 . . . . 1 .
1 . . . 1 . . . . 1 . . . 1 . . .
. . . 1 . 1 . 1 . . . . . 1 . . .
. 1 . . . 1 1 . . 1 . . . . . . .
. . 1 . . . 1 1 . . . . 1 . . . .
. . . . 1 1 . . 1 . . . 1 . . 1 .
. . 1 . . . . . 1 1 1 . . . . . .
. . . . 1 . . 1 . . 1 1 . . . . .
. . . . . . 1 . 1 . . 1 . 1 . . .
. . . 1 . . . . . 1 . 1 1 . 1 . .
. 1 . . . . . . . . 1 . 1 1 . . 1
. . 1 . . . . . . . . . . 1 1 1 .
. . . . 1 . 1 . . . . . . . 1 . 1
. . . . . . . 1 . 1 . . . . . 1 1
The solver chokes on anything larger than a 6x6 grid, but I've had good luck adding one row at a time to a smaller solution.edit: He mentions that he already found a size 74 set in one of the linked pdfs. http://www.cs.umd.edu/~gasarch/BLOGPAPERS/17x17.pdf