"Best" is relative and isn't the right question. Instead, you need to know your options.

Option 1: Assume no flexibilities and treat it as a normal WRC calc as shown.

Pros: Fast. Less expensive.
Cons: Potentially wildly inaccurate, both over-conservatively and under-conservatively.

Option 2: Calculate flexibilities via WRC, apply them, and treat it as a normal WRC calc as shown.

Pros: Still fast. Equally less expensive. Less over-conservative.
Cons: Still potentially wildly inaccurate, and still and maybe more under-conservative.

Option 3: Assume no flexibilities, plug loads into facilitated FEA software and calculate stresses.

Pros: Less fast. More expensive. Unlikely to be under-conservative see commentary under option 4.
Cons: Over-conservative, but less so that option 2 and 3. See commentary under option 4.

Option 4: Calculate flexibilities in canned FEA software. Plug into CAESAR. Pull loads out of CAESAR. Calculate stresses.

Pros: Even less fast. Confusion as to what proper flexibilities should be. More expensive. Most accurate thus far.
Cons: Is still a gross simplification of reality. Imagine flexibilities being provided as kx, ky, kz, krx, kry, and krz. These are calculated indiivudally as the average spring value over dx=0 to dx=fail. dy=0 to dy=fail. dz=0 to dz=fail. etc.

In reality, this spring value isn't going to be linear. In reality they will be a function of not only their own respective displacement, but also displacements in other directions. E.G. kx = f(dx, dy, dz, drx, dry, drz).

Not only displacements, but also temperature. Not only temperature, but also pressure.

But those options are not available to us.

Option 5: Full blown FEA of all things to be considered.
Pros: Most accurate.
Cons: Most Expensive. Also usually requires splitting the load with simplified FEA (e.g. CAESAR). Also hardest to ensure that you've captured all potential combinations.

Option 6: AI calculation of all said scenarios.
Pros: Most accurate.
Cons: Doesn't exist.