The question the article is about is 'can we deduce that a condition is discrete when we cannot measure it directly, only via symptoms'. If we could measure it directly - than we would obviously get a discrete distribution - but lets assume we we cannot do it. For example lets assume that we don't have the hi-tech flu tests and we can only measure temperature, frequency of cough, and some more subjective stuff like muscle and body aches, sore throat etc. These symptoms are not discrete - if we measure them in the whole population we shell not get any discrete distribution. So the fact that we can still deduce that there is a discrete condition underlying these continuos distributions is surprising - but apparently there are methods for that and one of them is called MAXCOV and it is briefly described in the article.
Also - what is the author conflating - can you explain what you mean by that?