The matter you mention is rather complex and highly dependent on details- the steel composition, quenching temperature and accumulated hot pre-strain (in case of no-rex austenite). Relevant dilatometric experiments are also helpful to get sure in BOTH CASES that  you have passed the transformation in the (same) martensitic temperature range. With no such information several assumptions would be equally acceptable as, for instance:

1) Deformed (non-rex) state of austenite promotes bainitic transformation and low-T lath bainite is often similar (morphologically) to martensite. Meanwhile statistics of (inter-variant) disorientations  within a packet are quite different for these species of alpha-phase. See, in particular Morito-2015 (DOI: 10.1016/j.matpr.2015.07.430) and closely related papers by Furuhara and Miyamoto.

2) Evidently REXed austenite was quenched from higher T where a lot of annealing twins could form in appearing equilibrium gains.  Related 60 deg boundaries inevitably should be inherited by a transformed structure.

Dear Ali Chabok,

Indeed martensitic or diffusionless transformations are complex and simultaneously  of highly technological interest. Being retired I cannot claim to be in front of the research. Nevertheless, your question puzzled me and I dare to make some preliminary remarks.

0. I suppose that your specimens of study are all of the same type (composition) of hypo-eutectic low-alloy carbon steel: in essence an  Fe-C alloy. Then austenitization usually takes place in the region given by some 50 centigrades above the A3-line. The reycrystallization as you mentioned is then performed by passing this line. The grain size of the gamma-phase is then solely determined by the time of austinitization. 

1. Porter and Easterling (Phase transformations in Metals and Alloys, Van Nostrand 1980) state in their Table 6.2 p.390 that for common low-carbon Fe-C alloys two different gamma-phase habit planes may be active. with different defect structures.

2. Different austenite grain sizes might invoke different orientation distributions of the grains concerned, e.g. via different gamma-phase habit planes, giving rise to different grain orientations before and after the martensitic transformation.

3. As you surely will know better than me there are a lot of studies on this subject, varying from nearly pure mathematics to nearly pure phenomena. Nevertheless I would refer to: 

Physical Metallurgy by Peter Haasen, Cambridge University Press, Cambridge UK, 1978, Chapter 13, p. 303.

But perhaps you tend to find this reference 'outdated'?

I wish you a lot of success with your studies!

Dear Chabok,

Your question is of the type that I may only indicate a relevant point to the underlying answer. So, admittedly, mine will be a comment rather than a formal answer.

I am assuming that when you mention 'fraction of high angle boundaries in martensite' you were referring to the angle between martensitic plates.

Unrecrystallized austenite, if you were quick enough to quench after hot deformation, should contain greater amount of dislocations prior to martensitic transformation as compared to the recrystallized austenite. Although the steel composition is the same in the two cases the austenite containing higher dislocation density (unrecrystallized) would have a higher Ms temperature. As the nucleation of martensite easier in such a high-dislocation-density-austenite packet-size (as you named them) can become somewhat smaller. Intuitively speaking, the angular difference between the martensitic plates may also be less for the same reason.