In fact it seems very stange that refinement. The alloy is cubic, right? If so, the wrong position of the calculated peak cannot be due to a mismatch of a lattice parameter. In your pattern there is only the 200 that pertains to the h00 reflections. Have you thought a possible error in the chemical composition of the steel? Do you varied the occupation of the cationic sites? Although usually the peak position depends mainly on the lattice parameter and on microstructural defects.

Did you already checked other materials as well?

I assume that your sample is perhaps slightly out of focus. This causes a shift of all peaks but the refinement will compensate this by a slightly change of lattice parameters. Unfortunately the difference plot shows then over two theta profiles which changes from reflection to reflection,(but systematically.

BTW: Is GSAS really using [hkl] - square brackets - for the reflection description? Correct would be the use of hkl (without any brackets) as given by Marcella.

The reflection shown in the figure does not seem to be a single peak. It looks like it consists of a double peak structure. You may see that by selecting a region around this peak profile and trying to fit with a single peak, and then a double peak (which I expect would give much better results). For that purpose you may use a software such as Origin (just to check the peak profile).

Thank you all for your answers. 

@Marcella: As you said the lattice is cubic and what is strange is that GSAS shows where the location of calculated peak is which matches the obtained profile (two little vertical lines under the peak in second figure) but other parameters ( Uiso, texture) seem to move this peak in the wrong direction while other peaks are very well modeled. There are some elements (C, Mn, Si, Mo) dissolved in the lattice in substitutional and interstitial lattice points I don't know how I can introduce them to GSAS. The lattice how ever stays cubic dominated by Fe atoms.

@Gert: We already know that more that 95% of the microstructure is made of ferrite phase which is BCC corresponding to the observed pattern. I wrote the labels myself, GSAS is innocent :D. 

@Sami: The pattern is taken using dual beam Cu and I have used it as an input parameter for refinement. 

Neither texture nor Debye-Waller factors may cause a shift of peaks. They only influence the intensity of reflections which can virtually shift a peak, but deconvolution  in GSAS will show you that this only virtually happens. A shift can only result by lattice parameter change, symmetry change, or misalignment of the sample (i.e. violation of the Bragg-Brentano focusing). Often you can see the effect on the residual intensity plot where intensity is missing. Texture and Debye-Waller factors only cause a "ghost" plot of your diffractogram, i,e, intensity is missing or too much at the same position. This will also influenced by the alloying elements if they are statistically distributed (substitution of Fe). Elements as interstitialy will "deform" the lattice and causes deviations from the assumed periodicity of the lattice. In consequence peaks appear broader than for perfect crystals (cf. martensite reflections will always appear broader than ferrite). S-shaped difference profiles means, the positions are incorrect. If they have the same character (left too much, right too small) it is usually the lattice parameter, but GSAS will correct this automatically. Problematic are shift which do not correlate with the lattice parameters. Then GSAS and other programs try their best to reduce the difference plot, i.e. misalignments will be compensated by a change of lattice parameters in order to reduce the least square of the deifference plot. You could enable a symmetry reduction to tetragonal (proposed by Sami) but I am not sure why this should happen. Nevertheless, it seems so that the shown asymmetry is more than the K alpha doublet can explain. Are you observing bulk? How do you prepare your sample to make it flat? Did you grind it and generated residual stress which can influence certain reflections in a different way? Or do you measure at non-flat surfaces?

From a quick inspection of pattern and fit, it looks as some type of planar defects is present in the BBC structure. If you read chapter 13 of the book X-ray diffraction (B.E. Warren) you will find that this kind of faultings leads to anisotropic broadening and asymmetry on Bragg peaks as well as individual displacement which could fit your pattern. Nevertheless, before you apply this formulation (available in several analysis programs based on whole-pattern fittin) you have to be sure about discard all mentioned factors as missalignemts, lower symmetry, ... 

@ Tanmoy Paul: As far as I know serious asymmetry influences appear preferably at low 2theta angles, i.e. the asymmetry should be clearly higher for the 110 but there you don't see anything. It does not make sense to activate any parameter only because it is possible in a software. The better  way is to do it vice versa, at least from my point of view. It is mathematically clear that you will get a better fit the more parameter you are activating for refinement. The question is only, what do they still tell you?