dilog 'silent' error at NLO

Hi,

I am running one channel of three tops at NLO using the following command lines:
--> import model loop_qcd_qed_sm_Gmu-no_widths
--> generate p p > t t~ t~ t~ w+ [QCD QED]
I do fixed order computation aiming to look at the different contributions from each order. I treat the overlap from tttt, t>wb and the process converges and produces the expected LO1 and NLO1 cross-sections, but then beyond those orders, the other contributions are either zero or just random.

I noticed a long list of dilog errors in the log file of the run, one line of which looks like this:
------- iteration 3
 Update # PS points (even_rn): 29120 --> 24576
 ERROR in OneLOop dilog_r: |x|,iph = .1000000000000000000E+01, 2 , returning 0

The code never complained about this in run time, and so I am claiming the run failed silently and this relates to the random contributions I get beyond LO1 and NLO1.
I am assuming it is an error in the tensor reduction, but I am not sure how to fix or even debug this in mg5, may you please help?

Best,
Hesham