If I read the description correctly, the glass was tilted to 15, 30, and 45 degrees, but I may be making the same assumption as JimH.
As I understand it, basically the test appears to be largely two dimensional. Notice how 15 degrees was denoted as sharply sloping. I take that to mean that 15 degrees was from horizontal and so 90 degrees would be vertical.
Neat test for a benchmark start. I assume that results would only improve as the glass became even more vertical, as it is in many types of vehicles.
Since the test apparently was only done shooting straight on at the glass, as if the shooter were in the road and was shooting at an approaching vehicle, the next round of tests should add the third dimension and have the shooter shooting from various degrees away from the center line or being directly in front of the vehicle. No doubt that the combination of windshield slope in conjunction with the degrees from center line will show that the extra added angle from center will further reduce potential penetration.
Cool testing idea. I hope that ajp3jeh will follow up with more testing!