I doubt it's the sensor's fault. When you use the electromagnetic type partial discharge measurement devices they go nuts around the commutators of DC motors because by nature the commutator is continuously sparking at levels far above that partial discharge meters are meant to detect. You have to switch to another measurement technology to overcome this. My suspicion is that the SKF meter (we have one in the shop) is similar in nature...it gets swamped by the extreme sparking RF produced by normal operation of a commutator. My suggestion is to go back to the traditional measurement method.
I've used that one as well as the pickup made by Aegis as well as just taking an ordinary chipping hammer with the brush on the back. We never got anything resembling representative data (same readings) from the SKF meter. It sits on the shelf and collects dust. We've pretty much given up on it, whether following SKF's instructions to the letter or just moving it around and trying different things to get a reading. Right now we keep it around in case a customer ever asks and bring it along just to show them that it doesn't work whenever we get a call on shaft voltages.
I've also tested a good high speed multimeter "peak" function against the oscilloscope method and gotten similar results (same millivolt readings). I hate to say it because they're expensive but the Fluke Scopemeters do a great job as far as you can just hit "Auto" and maybe fiddle with the adjustments a little more to either fix whatever the Auto function missed or verify that you are getting the correct reading, usually just zooming in on the peak signal over what the Auto function finds which is usually 2-3 cycles of the underlying waveform. If you have the right multimeter and the right "peak" function it works just as good as the oscilloscope except you can't hit the "print" button and you don't have a nice visual reference to show that you got the right readings. This is important when you are looking for something under 1 V which might be in the range of 0-10 mV in the vast majority of cases (don't want to miss a real problem). And if you don't have an oscilloscope and you haven't detected real bearing shaft voltages so that you don't have anything to try it out on, you don't have any way to validate your multimeter, but I can tell you that Fluke 87 V's are fast enough. Haven't had a chance to try it out again on say a much cheaper Fluke 117 or an Amprobe 300 series so unfortunately I can't help you out on the meter cost other than to suggest that although I haven't used them, there are some really inexpensive digital oscilloscopes out there that easily beat the specs of the one that comes in yellow plastic with a black border (that is trademarked) from some better brands that I wouldn't hesitate to use. We have the Fluke branded one. If we didn't, I'd buy something else.