Guess it's too late to change the past but how does your shutter tester work? If it's a light sensitive cell aimed at a light, then a penumbra could "start" the time count early and "end" it late, resulting in "longer" measured times than effective.

I'd rely on a sensitometric test, exposing frames at different times and compare 1/30th f/16 to 1/60th f/11 for example, working your way up to 1/1000.