I figured it out. Still testing at 128 duty cycle.
If I hook it up to ground, and the + side to pin 3, it reads perfectly fine (same as transistor).
If I hook the meter up to 5V, and the ground side to pin 3, then the meter reads 4V for some reason.
I thought that a PWM pin would sink the same amount of current as it would source, but apparently that is not the case. If the meter reads a higher voltage when the pin is set up to sink current than when it's set up to source it, then I guess that means the pin doesn't sink as much current as it will source. I think.