Here is what I would suggest.
From what I understand you have PDFs from two sources.
For one source, they load successfully but for the other they fail to load.
1. Take one sample PDF from each source.
2. Run arspdump on both the files. Ref. -
http://publib.boulder.ibm.com/infocenter/cmod/v8r3m0/index.jsp?topic=%2Fcom.ibm.ondemand.mp.doc%2Fars1d171399.htm3. Then open the output of the arspdump and search for the Trigger, Field, etc that you defined using the graphical indexer.
4. Compare the UL, LR that you get with the values below.
I suppose that for the files that load successfully, these values will match.
Whereas for the other it would not.
TRIGGER1=UL(0.15,10.53),LR(0.42,11.49),*,'TOPAS-BEL-MET'
FIELD1=UL(0.29,9.34),LR(0.53,9.72),0,(TRIGGER=1,BASE=0)
FIELD2=UL(0.43,9.31),LR(0.67,9.68),0,(TRIGGER=1,BASE=0)
FIELD3=UL(0.86,9.15),LR(1.15,9.68),0,(TRIGGER=1,BASE=0)
FIELD4=UL(1.17,8.92),LR(1.42,9.67),0,(TRIGGER=1,BASE=0)
It may also happen that due to the modification being done in the PDFs, the TRIGGERS and FIELD coordinates are getting off than the original.
It could also happen that there is additional string in the box that has TRIGGER or the fields.
I have been through similar issue where we were using iText to modify the PDFs to insert hidden triggers, etc.
So you may end up creating a separate CMOD Application for one set of PDFs and another for the set that fials to load. Offcourse with different coordinates.
Hope this helps.
Cheers
Pankaj.