ADE Assembler, parametric: (only) first simulation fails

dear cadence forum,
I have recently moved to virtuoso IC6-1.8-64b.500.14, and I am seeing a weird effect that I was not experiencing in older versions.
I wonder if anybody else has seen this effect and/or has some suggestion to solve it.

here is the issue:
if I run a parametric simulation in Assembler, the first parametric step always fails, while the the further steps work as I would expect.
so, for example, if I define a variable Nph and run a parametric with Nph as 0 30 70, the simulation with Nph=0 fails while the simulations with Nph=30 and Nph=70 run correctly

I am reasonably sure that this has not to do with the parameter value because:
- if I run a non-parametric simulation in that condition (Nph=0), it runs correctly
- if I run a parametric simulation in which I introduce another point at the beginning of the list (e.g. -10 0 30 70), then the simulation with Nph=-10 fails while the simulations with Nph=0, Nph=30 and Nph=70 run correctly
 
why does the first point of the parametric fail?

CIW reports ERROR (ASSEMBLER-5031) (The simulator process returned a non-zero exit code, indicating failure), and tells me to check the spectre.out file
the spectre.out file that CIW suggests me to check does not exist.

the Job.log file in that directory tells me that:
\o ERROR (ADE-3036): Errors encountered during simulation. The simulator run log has not been generated.
\o         Possible cause could be an invalid command line option for the version of the simulator
\o         you are running. Choose Setup->Environment and verify that the command line options
\o         specified in the userCmdLineOption field are supported for the simulator.
this however does not help, because a simulation with the same Setup->Environment settings, using the same parameter value, works perfectly fine if I run it as a single simulation (not as a parametric simulation)

to add to my puzzlement, the Job.log file in that directory tells me that:
\o *Info*    The simulation mode in Spectre has been changed from
\o           'interactive' to 'batch' mode because  simulation errors
\o           found..
Only, this seems to be a lie, because if I check the simulation mode in Spectre with the command
envGetVal("spectre.envOpts" "controlMode")
I see that it has remained as "interactive"

BTW I get the same results (first simulation not working, the others working ok) if I change simulation mode in Spectre to batch mode before starting the parametric simulation

as today, I am working the problem around just by "tricking" the system: I add a first "dummy" step to the parametric list (that will fail), so that the following steps (which are the ones I am really interested into) will be executed correctly.
but I wonder if there is a a better way to avoid the issue completely