AVR Freaks

First, I want to commend you for actually doing the experiment. But, along with that, there IS a caution. It is hard, I mean REALLY hard, to make meaningful measurements in this situation. Been there, done that, too many times. 

It is not at all surprising that the 10R has the least ripple nor is it surprising that the straight 10uH inductor has the most. That is because the combination of the inductor and the bypass capacitor make a parallel resonant circuit as seen from the load (the MCU). Ferrite suppressor beads add an equivalent parallel resistance to the inductor and this tends to damp the ringing. 

Now, to the question of whether or not it is "good" or "bad". The answer is an unequivocal "it all depends". It might have a poor effect if you are using the comparator or ADC. Most of the rest of the time, it is pretty much "who cares?".

So, why go to the effort of even adding isolation components? Well, especially if you are trying to meet radiated emissions limits for CE or FCC certification, it is a REALLY big deal. That is because the MCU creates power supply current spikes on every clock edge. These current spikes can be quite large (my long-ago back of the envelope computation suggested 100mA) and very narrow (equal to the logic level rise time, which is in the range of 5ns to 10ns). These large and narrow spikes can cause traces through which those currents flow to radiate at much higher frequencies than any clock in the system. Those radiated signals can make it impossible to meet radiated emissions limits. So, what you are doing by adding a lossy low-pass filter is to lower both the amplitude and the frequency of the currents into the rest of the system. You pay for that, of course. Part of the "pay" is more ripple on the load (MCU) side. 

Part of the challenge, here, is that those high amplitude, very fast spikes, are REALLY hard to see with an oscilloscope. First, you need a scope with at least 200MHz bandwidth. Then you need a GOOD probe. Then you need to connect the probe in a way that gives you accurate signals, and that usually means zero length ground connection at the probe tip. If you don't do all of these things, then you simply won't see the fast spikes (or you will see a false representation of them) and you won't understand what they are really doing in your system. 

Now, this does not fully answer the original question. But, I do hope that it helps you to understand what you saw and helps you to determine whether or not it is a problem. 

If it were me, I would use a cap larger than 100nf and I would use any of the ferrite suppressor beads, and be done with it.

Jim