In contrast to LR parsers, LL parsers are quite easy to write and maintain by hand (recursive descent, parser combinators, etc.), and the nature of such hand-written parsers makes it comparatively easy to step outside theoretical bounds and implement context-sensitive portions of the grammar. Such parsers have some other benefits, too, like being easier to debug and making it easier to provide useful error messages (but see Menhir for OCaml, for example, which offers some improvements in these areas over traditional LR parser generators).
What Appel has said is true—it's rare to encounter hand-written LR(1) or LALR(1) parsers in practice. However, many practical languages are easy enough to parse with hand-written top-down parsers that it's quite common to encounter hand-written LL(1)-ish parsers in practice.