That "any prop" part is the killer. Most people get less thrust from trying that:
My experiments suggest only a a max of 20% improvement is possible:
It should spin slower (because more of the prop is producing lift, instead of stuck in the vortex at the tips) for the same power.
The "well designed" part of your question is also non-trivial - there's no such thing if you want the prop to perform optimally across a range of airspeeds - the duct has to be designed for a specific operating condition, so it will always make things worse at "off conditions" - which will usually be more worse than the prop without the duct.
The entire point of the duct is to reduce tip vortices, and the entire point of the tip region of "any propeller" is also to reduce vortices, so using both at once is not going to work properly, if at all (see video above - in real life, it does NOT work). The tips of a duct fan need to be totally different shape to the tips of "any propeller".
I question the accuracy of the paper mentioned: he shows that reduced diffuser angle improves thrust (p. 51 - but totally fails to consider Bernoulli for that, and never even tried angles less than 0 degrees), and his results and diagrams are almost all based on large diffuser angles, which is the opposite to Bernoulli, common sense, and modern practice.
In my relatively vast experience in this field, almost all CFD models and predictions are false (papers are usually written by students, none of whom understand modern ML techniques, especially the "time traveller conundrum" as it applies to CFD validation). Long story short - they do not realize that they're "training" their math to match the observations [almost always of poor experiments, and always of a tiny sample], instead of what the math is supposed to be - a prediction, for more than just the sample observed. Bottom line - if it's not something real that was measured, you can't trust the numbers.