SNMP - Interface statistics query + In/Out bits with Total Bandwidth graph type + traffic over 113something Mbps = graph rollover

Hi,
using Cacti 0.8.7i;

the topic more or less summarizes what I am experiencing;

The SNMP query of Interface Statistics used came from the Generic SNMP-enabled Host
Host Template.

Whether I select the In/Out Bits with Total Bandwidth as normal or with 64bit counters,
it doesn't change what the graphs show: when the (5 min avg) throughput goes over 
about 113 Mbps, the aforementioned 113Mbps is being substracted from the value returned
and so the graph never shows more than 113Mbps.
That wouldn't have bothered me much... before the link this is supposed to graph, got
upgraded to 400Mbps (it was 185Mbps before, so it was clearly visible, where the rollover
had happened...) Now it can rollover twice or thrice, it is itching...

What I tried to do, was to cause the graphs to not use rigid boundaries and 
set Upper Limit to 1000000 (not using per-graph value) while still Auto-scale being on;
(in graph templates);

Also made the data template for Interface-traffic data source, Maximum Value to 'U',
both on traffic_in and traffic_out;

I don't know maybe there was some action required upon these changes?
Still I accept that the historical data will not be amended that way, but the real question,
is - how can I find out the values being graphed, starting from result of SNMP query, through
data source, to what's fed into graph/rrdtool? To find the most likely culprit.

Could it be, that the queried device returns 32bit values and that is the real problem ?

Kind Regards,
Lukasz


------------------------------------------------------------------------------
Free Next-Gen Firewall Hardware Offer
Buy your Sophos next-gen firewall before the end March 2013 
and get the hardware for free! Learn more.
http://p.sf.net/sfu/sophos-d2d-feb