If I understand the question correctly, I think the answer is: there’s no difference, just two different methods that achieve the same result. i.e. grouping two or more tasks will create a new action group with two subtasks; creating two “child tasks” (probably better referred to as “subtasks” in OmniFocus-land) would result in exactly the same action group.
At the risk of answering a rhetorical question: I definitely use both ways! I think it depends on whether you first have the tasks and then decide that, actually, they should be included in the one group; or, alternatively, start with the main task knowing you need to add subtasks.
The difference is that one is tasks in a project, but sometimes projects are far more complex, so the use of Sub Tasks allows you to break that project down into stages with different parameters, e.g. some stages may require tasks completing in order and others, it doesn’t matter.
I use sub tasks with templates to break projects down into smaller sections to show progress.
It’s amazing how many answers people give to rhetorical questions, isn’t it?
The way I see it is that DT allows you to categorize/subcategorize things on pretty much infinite levels. I’ve never tried more than 3 or 4, but I’m not aware (and I can’t imagine I’ll ever practically find) a limit.
The problem is when you find yourself with half a dozen things that you’ve already created that should be part of a sub-task, sub-project, etc.
Without a separate “group” command, you’d have to create a new task, select those items, and drag those items under the heading manually. “Group” does that for you with a single command.
No matter how you create them though, all DT is is a collection of parent/child relationships.