First, there's what I think is an important distinction of idiomatic programming in Scheme, for which the first example in the article gives the wrong idea, with a noble lie: that Java code implies an immediate return, but the Scheme code does not.
Were that Scheme expression not in a tail position of a procedure (which it probably wouldn't be), it wouldn't have effect analogous to that of the Java code. And a one-armed `if` makes little sense in a tail position in modern Scheme, and looks erroneous.
Scheme is a full imperative algorithmic language, and doesn't have to be used in a pure-functional way, but idiomatic programming in Scheme tends to lean towards functional when practical. You almost never do premature returns quite like that in Scheme, for example. (One can do most any memory-safe thing one wants, including the less-idiomatic and even the outrageously insane, but hopefully over time one learns the idiomatic ways, rather than overuses/misuses less-idiomatic features, like leaning on a crutch they don't need and that's actually harmful to them.)
Separate from that, another thing possibly misleading about this article is that their Scheme implementation doesn't do what's sometimes called "proper implementation of tail calls" (which would be required to be a real Scheme). Idiomatic Scheme will tend to use tail calls heavily, with impunity, which would not be the case if a Scheme implementation let such calls exhaust resources.