PDF Extraction never the same format

Hi, I just started trying your pdf extraction api for a client, it's seems to be working fine but I am having issues with how to get the needed data while all the PDFs are visually exactly the same. Therefore, the paths, rows and index are never the same for the same elements. 

Here is an exemple : 

For the first invoice PDF, if I am trying to get the first "Cat. No." :

Path : //Document/Sect/Table/TR[7]/TD[3]/P

Index : 68

Row count : 6 

For the second invoice, first "Cat. No." :

Path : //Document/Sect/Table[2]/TR[2]/TD[3]/P

Index : 60

Row count : 5

Am I doing something wrong or is the API not precise enough?

Thank you for your help!