Hi Jose,
> I cannot see ITL waits on these tables as well.
Well i hope so, if you have 100 ITL entries ... otherwise you would have a really high parallel DML load 
> For updates is very clear for me that it can happens but I'm not sure with INSERTS. What do you think?
An INSERT statement needs an ITL slot as well (and if several different sessions are inserting data into the same block - several ITL slots are needed). You can also have deadlocks by INSERT statements (just think about foreign keys, which are not used in SAP environments, but you can have such issues with unique values as well).
> Regarding the reduction, last weekend we have reduced the INITRANS from 100 to 1 and reorg the tables in a test environment but the result hasn't been the expected.
Then maybe some other influencing factors kick in or your expectations were based on some wrong facts.
> Honestly is very difficult to found something with block dumps because these tables have 64642928 and 5352312 blocks and I don't know which blocks were updated.
Well this is pretty easy to determine - just use the pseudo columns ora_rowscn and rowid to get the (latest) updated blocks and crosscheck a few of these. This should be just a tiny average cross check of several blocks and not a scientific result (for which ora_rowscn is not good at all).
Regards
Stefan