If the shutter speed of the camera is close to the power frequency being applied to the tube (or an even multiple) then it becomes possible to capture it during a point where the ac waveform is rising or falling and potentially be able to effectively "stop" the visual effect. This is why you can sometimes get part of a sign to appear on and part off in a single photo even though visually it all appears to be illuminated. Add to this that the cameras have a wider dynamic range than what you can see with the naked eye and you can easily create a situation where things will appear different. (You could, for example, be seeing a situation where the phosphor is emitting on the down cycle while the other is closer to the peak of the waveform.) At least that's a good guess without knowing any really specific information on the types/number of transformers and how the load was divided up. To mitigate this, try using a different shutter speed and aperture combination or even a different speed of film or iso setting.