The "Network Data Transfer Rate" is the result of a simple division; the
amount of data moved divided by the time it took.  When the denominator
(time it took) is very small in absolute terms, the division can result in
wild numbers.  A good rule of thumb is that a valid throughput calculation
should be based on a total "Data Transfer TIme" of 2-3 minutes or more.  Of
course, you don't have any control over this unless you're doing a
controlled study.  But it turns out that in real controlled studies of ADSM
performance, an effort is made to move an amount of data which takes about
this much time.  I've seen network throughput well above 10,000 KB/sec also,
but the Data Transfer Time was only few seconds or so.  When the time was
several minutes, the throughput dropped back down into the 5000 KB/sec range
on a network like yours, which fits more with reality.

In this case, the "Data Transfer Time" is .11 seconds, which is very small.
Way too small to give a number which is valid.  For example, if the transfer
time was a mere 1/100 of a second different, or .10 seconds, then the
quotient would change by almost 10%!.  You can ignore the 35Mb/sec as just
an artifact of the way the statistics are presented, when the Data Transfer
Time is so small.