Cacti shows different values than rrdtool command given in graph debug mode

Hi all,

I am trying to calculate the various values Cacti does when graphing a
network device interface (average, maximum, etc). Here is the graph
debug for an existing graph in Cacti;
http://oi45.tinypic.com/hsko3n.jpg (I have blanked out the device name
and IP etc)

I wanted to calculate the details on the command line, so I took the
rrd command run by Cacti and made a small bash script out of it, just
as below:

$cat rrdtest.sh
#!/bin/bash

rrdtool graph -f '' \
--start=$1 \
--end=$2 \
/dev/null \
DEF:a="$3":traffic_in:AVERAGE \
DEF:b="$3":traffic_out:AVERAGE \
CDEF:cdefa=a,8,* \
CDEF:cdeff=b,8,* \
PRINT:cdefa:LAST:" Current\:%8.2lf %s"  \
PRINT:cdefa:AVERAGE:"Average\:%8.2lf %s"  \
PRINT:cdefa:MAX:"Maximum\:%8.2lf %s"  \
PRINT:cdeff:LAST:"Current\:%8.2lf %s"  \
PRINT:cdeff:AVERAGE:"Average\:%8.2lf %s"  \
PRINT:cdeff:MAX:"Maximum\:%8.2lf %s"

$./rrdtest.sh 1357746014 1360424414 /var/lib/cacti/rra/XXXXXXX.rrd

Current:    5.24 k
Average:  153.55 k
Maximum:   13.65 M
Current:    5.06 k
Average:  158.57 k
Maximum:    6.07 M


As you can see from the link to that graph debug, none of those values
match the values in the graph yet I have pulled them from the same rrd
file using the same commands Cacti showed me. I have repeated this
with various rrd files for various host types (servers, routers,
firewalls). I assume that perhaps the graph template is changing them
slightly (although I can't see anywhere in the graph template settings
that would cause the values to be different!).

I have exported the graph template to XML and attached it to this
email, but just in case this mailing list doesn't allow attachments I
have pasted it here: http://pastebin.com/embed_js.php?i=GQMYZi8A

Why are the values displayed by Cacti different? They are only
*slightly* different? They aren't *8 higher or /8 lower, so it's not a
bits to bytes conversation. They aren't a "normal" ratio like half or
double either. They are about 1% different, how can they be *just
slightly* different?

If anyone can help me with this I would be greatly appreciative,
otherwise I shall have to work on the assumption that Cacti is
producing invalid values.

Many thanks,
James.

------------------------------------------------------------------------------
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