Keep in mind that Guinier fitting deals with only the lowest-q data (smallest scattering angles).  You state that your curves have "comparable SAXS profiles," but if you are comparing them over the whole range of q-values you measured (from essentially zero out to maybe 0.3 or 0.4 inverse Å) then this might obscure significant differences in the lowest-q data that are used for Guinier analysis.  Does the Guinier plot show evidence of interparticle interactions (i.e., aggregation or repulsion)?  These will manifest themselves by a systematic upward curvature as q gets smaller in the case of aggregation and a systematic downward curvature with decreasing q in the case of interparticle repulsions.  In your case, it could be that curvature in the Guinier region is making it difficult to identify a linear Guinier region at all (because there is no such linear region!).  If that is the case, my opinion is that it would be better to spend your time optimizing your sample (concentration, buffer conditions) to minimize the offending aggregation or repulsion and then collect higher quality data rather than spending your time agonizing over Guinier fits of data that are not truly suitable for Guinier analysis.  Finally, if you can manually identify a linear region in your Guinier plot, make sure you are not including data points that extend beyond qRg of about 1.3.  Beyond that, you are no longer satisfying the assumptions that go into Guinier analysis.  If any of the early points of your data are corrupted for whatever reason, remember also that there is a minimum q-value you must sample in order to analyze the size and shape of your scattering particles.  q(min) must be less than or equal to pi / Dmax, where Dmax is the largest interatomic distance in your scattering particle.  You mention you're studying bilayers; if these particles are large bilayers such as vesicles, you might have a large value of Dmax, which unfortunately means that you might have to make very exact measurements of the lowest q data, which can sometimes be difficult in practice.  If the data really do look good (linear Guinier region, suitable data points from a reasonable range of q-values available, etc), I have always found it possible to do manual Guinier fitting, even if the automatic Guinier fitting algorithms are less than satisfactory on the first try.  Simply adjust the position of the first and last point in your Guinier fit systematically until you get a reasonable result.  Finally, always be sure to look at the residuals of your Guinier fit; if they show systematic deviations from zero, it's another indication that there are problems with the underlying data.  Good luck!

Thank you very much Dr. Whitley for preparing such a comprehensive and lucid answer to my inquiry. Your description is very reasonable and helpful to my purpose.  

Thank you very much Artur for your answer. I will give a try to this also. I have been using Origin Lab for data analysis. The SAXS gadget on it didn't perform well. I'll try Igor Pro too.