Well, addressing the concern directly i.e,

why computed code stress is half of the theoretical computed stress and similarly why the theoretical allowable is not halved accordingly as it remains the same as is derived based on yield limits, lets have this comparison:

(Commercial pipe) (Polished homogeneous tube) (Theoretically calculation)

SIF 2 1 1

Or

SIF 1 ½ ½

As the commercial pipes are used practically and CODES being meant to simplify the processes as simple as possible, have judiciously enforced the half of the theoretically calculated SIF values.

Actually, when the CODE is dealing with the commercial, practical conditions, it needs to have THOSE DATAS (SIF’s) directly put in instead of entering into the loop of theoretical or perfect/ idealized conditions and back calculating the results. This same is referenced by Sir L.C. Peng in his coveted work when he says “By using the commercial pipe with an unpolished girth weld as the basis, the code SIFs as given by eq. (3.11) & (3.12) are only one half of the theoretical SIFs. THE ADOPTION OF THIS BASIS IS MAINLY ATTRIBUTED TO PRACTICALITY.”

However, as the above things are mentioned in section 3.4 of the book, Sir L.C. Peng CONTINUES to write in section 3.5 of the same saying:
“the the stress allowable derivation of non-yeilding benchmark stress range REQUIRE FURTHER JUSTIFICATION before being used”.

Here it’s been clearly stated that why the STRESS ALLOWABLES derived from the yeild limits are not being factored.
“The first thing….” adresses the discrepancy arised due to SIF but the “Other thing….” clearly emphasise the practical aspect of fatigue cycle.

He further adds that “with f=1 at 7000 cycles, the allowable stress has already reached the benchmark stress limit. Any stress beyond that might produce gross yielding in the system, thus invalidating the elastic analysis."

Hence as pointed out by Durga, "I think that is the reason the benchmark stress range SEB is 2 Sy i.e. SEB = 1.5(Sc+Sh)",
half of code used SIF for COMMERCIAL PIPES is ONLY A factor which further JUSTIFIES the stress allowable derivation of non-yeilding benchmark stress range.

Hope this clears the doubt!