This is covered in the FAQ. Partly, I allow it because there are no parsing ambiguities with it, so why not. But moreover, depending on type name & variable names, intelligently beginning or ending with an alias can lead to extremely fluent APIs. "that(expected Bool)Equals(actual Bool)" vs "writeTo(Linkable output)". This freedom can lead to more readable APIs in any language, not just English.
Of course, I do see where the complaints come from, and a part of me fears the day that someone releases a wake linter which requires aliases to follow the type. If that were to happen I think I'd nix the idea in favor of unity in the community.