Adding a parallel 0.1 microfarad ceramic capacitor to the existing bulk output capacitor does nothing useful.
It may be needed for stability if feedback loop bandwidth is quite high. At least that was a suggestion from one application note which I don't remember. The idea was that bulk electrolites may be too "slow" for high-speed regulation.
I personally always include a 100nF on the output, but never measured the impact of this. Although, when I do measurements that can be affected by the power supply, I try to make power cable as short as possible, often connecting the test fixture directly to output bananas, in addition to a capacitor a the point of load. But the thing is, because of parallel resonanse between decoupling caps (and their parasitic inductances), it's impossible for me to predict if adding more capacitors helps or not. I often try to measure "before" and "after" and compare if there is any change and if I have enough time. Otherwise there is some uncertainty. Fortunately, I work with hundreds of kHz, so this is not that big issue.
So I what to suggest: do measurement to see what actually makes sense, and what is not. Dave has two good videos on the subject: 1) showing impact of mlcc and electrolyte caps as decoupling caps 2) in red pitaya video he showed that decoupling caps in parallel may cause problems.