The study used all the GIH samples and all the PJL. So just to correspond I doublechecked some divisions.
Which letters correspond to which numbers are strong inferences because I know how many samples are in each group of both divisions. I have the original data sets; both Zack Ajmal's divided set and the unadulterated 1000Genomes data set.
By the numbers:
GIH_1 = GIH_Patel in the Zack Ajmal Dataset. Exactly 67.
GIH_2 = GIH_ANI 1/2/3/4 = Gujarati A/B/C. Exactly 36.
PJL_1 = PJL_A minus 2 samples OR PJL D minus 4 samples.
PJL_2 = PJL_B/C/D plus 2 samples OR PJL A/B/C plus 4 samples.
PJL A 
33 

PJL 1 
31 


PJL B 
9 

PJL 2 
65 


PJL C 
19 





PJL D 
35 
















GIH ANI 1 
8 

GIH 1 
67 


GIH ANI 2 
8 

GIH 2 
36 


GIH ANI 3 
15 





GIH ANI 4 
5 





GIH PATEL 
67 

Corresponds exactly to the divisions. 







Every sample is used. No sample is unaccounted for.



EDIT: Originally, I wrote which ANI/PATEL group corresponds with which GIH_A/B/C/D. However, I need to doublecheck before I type that out.
This diagram and table lists n= for each sample set from the Sengupta et. al. study and their subsequent divisions. 
As for how the GIH_Patel group got its name; or why they are
assumed to be Patels is because when the GIH samples first came out they plotted like this:
So every single person who analyzed it, knew that there was at least 2 distinct groups. Zack Ajmal then compared that to
known Patels he had from his Harappa project, and the assumption became that these ~67 are Patels.
The rest of the samples then got divided by ANI, so you got 1/2/3/4.
This part is important. ORIGINALLY, Gujaratia was representing that group that clustered tightly. Gujaratib was those that DID NOT. Gujaratia gets renamed GIH_Patel AFTER Zack's analysis. All calculators at that time only had 2 division for Gujarati, A and B.
Subsequently, Gujarati A/B/C/D names were made, corresponding to ANI1/2/3/4 and Patel. Which is which I am not exactly sure. That I will have to dig more deeply into. However, the important thing is that the definition of Gujarati A on admix calculators flipped!