The client which my firm is working with wishes to use non-spec wall pipe that has a d/t ratio greater than 100 (D=914mm, t=7.94). This thin of wall is introducing a host of other problems which I won't go into.
Caesar gives a notice (warning) that "SIF computations for this fitting may be invalid." I've checked the values Caesar II reports against SIF values from FEA and they are indeed no longer accurate for long radius bends. I am trying to locate the section of the code which explicitly defines the d/t conditions for bends. Mandatory appendix D is silent on the d/t issue for bends. Could I just override the Caesar II values for SIF's and insert user defined SIF and flexibility factors? I would hazard a guess that the code doesn't allow this. I know the history about J.G. Merkle and his empirical testing to determine SIF values so I know there is a solid basis for the limit on d/t ratios but I need to have a reference I can point to in b31.1.

Check the "Notes" to Appendix D, that is where you'll find the limitation on the D/t ratio.

Yes you can override the "Code" values, just type in your "more applicable" SIF value in the input.

^Yes, note #13 defines the d/t < 100. Note #13 only applies to reducers and not bends though (B31.1 2007). The description for bends include notes 1,2,3,4,5 none of which call out a limitation on d/t ratio (unless I am mistaken). There must be a reason Caesar is flagging it though. In B31.3, note 1 (for bends) is more telling than its equilalent in in B31.1. "Stress intensifiation and flexibility factors...are for use in the absense of more directly applicable data...their validity has been demonstrated for d/t ratios < 100." Since both 31.1 and 31.3 are based off of BPVC, I would assume that d/t ratio is applicable to 31.1 even though it does not say it outright.
Also, would data obtained from FEA qualify as "more directly applicable data"? It sounds as though 31.3 is saying I can override the SIF and K values in Caesar. Its probably not a stretch to say that 31.1 allows this as well since they come from the same original code.

The client is 'detailed orientated' so I apologize if my questions seems rather trite.

Exactly, see Note #1 in B31.3.

Yes FEA data qualifies as "more applicable" data.

The Codes are a "minimum" set of guidelines. You are free to perform more exact analysis based on better data.