There's an option in the bend dialog to change elbow thickness. You may alternatively break the element "before" and "after" the bend so that you can have your elbow be a discrete element on its own, but that's tedious, but I suppose there are benefits:

1. You can visually review elbows (and therefore provide visual documentation) that the elbows are indeed different.
2. You can assign larger ODs and WTs, if this is the case. (Note that you'll also need to look into custom specifying elbow radius, too.)

There are also detriments:

1. More time to input. More risk of error.
2. More likely to encounter "do not specify node on bend" errors.
3. More node management.

In my experience, you shouldn't have to input SIFs (or flexibility factors) on bends in most normal cases. Exceptions include:

1. You have FEA results you want to incorporate, just for a more accurate answer.
2. You have a trunnion attachment, in which case you have to use FEA.
3. You have non-standard bends for your application that have their own SIFs and flexibility factors per the manufacturer.

The last one is a bit of an oddity. I've seen it in cases where manufacturers can't yet comply with new FRP standards (maybe they're working on it, maybe they aren't), and continue to re-purpose codes such as B31.3 for FRP.