The reason is because of their presentation.
We can see that Cosine is better for the similarity analysis since we want to know the distance as close as possible. But if we use it in the plot the readers will have a difficulty to see since the distance between terms are too close. Since the words are crossover one another more often rather than the Euclidean distance.
On the other hand, Euclidean shows a better presentation.
I drew three red circles to form the clusters of terms. The first circle is where data and analyst terms are located, follows by the next circle where the terms of experience are situated. Last order red circle which the bigger cluster shows the various terms to explain more about the job.
this is based on the 25 November 2020 Dataset