Hi Eric,
That's normal, I'm getting very similar values with my P6248 and TDS784A at 100 mV/div.
At 10 mV/div I'm getting around 10 mVpp, without the probe I get about 2.5 mVpp (that's the noise from the scope itself), so let's say the probe itself adds about 8 mVpp.
You have to remember this probe is meant to be used for signals up to 1.7 Vpp (in 1x mode), so ~10 mVpp of noise is perfectly acceptable and it should be just fine for most Ethernet testing.
Additionally, the probe has a bandwidth of 1.7 GHz; if we recall that noise goes up proportionally with bandwidth (assuming it's white noise), then its high bandwidth will inevitably lead to much higher noise levels than your typical 500 MHz passive probe.