using tc control command in bash script gives unexplainable results - bash

Question

I send frames in packets from a client to a server on the server I want to shape my traffic.
I use this script to control the traffic. First down to 80 kbit after 10 seconds down to 40 kbit. (I know this is ridiculously low, I usually use bigger values)
#!/bin/bash
datenrate=80
datenrate2=40
echo "setting datarate to ${datenrate}"
touch started.info
sudo tc qdisc del dev ens3 root
sudo tc qdisc add dev ens3 handle 1: root htb default 11
sudo tc class add dev ens3 parent 1: classid 1:1 htb rate ${datenrate}kbit
sudo tc class add dev ens3 parent 1:1 classid 1:11 htb rate ${datenrate}kbit
echo "worked"
MSECONDS=$(($(date +%s%N)/1000000))
STOPTIME=0
while :
do
STOPTIME=$((($(date +%s%N)/1000000) - $MSECONDS))
if [ $STOPTIME -ge 10000 ]
then
sudo tc qdisc del dev ens3 root
sudo tc qdisc add dev ens3 handle 1: root htb default 11
sudo tc class add dev ens3 parent 1: classid 1:1 htb rate ${datenrate2}kbit
sudo tc class add dev ens3 parent 1:1 classid 1:11 htb rate ${datenrate2}kbit
touch calledthrottle.info
break
fi
done
echo "10 sec over - setting up a datarate drop to ${datenrate2} kbit"
while :
do
STOPTIME=$((($(date +%s%N)/1000000) - $MSECONDS))
if [ $STOPTIME -ge 20000 ]
then
sudo tc qdisc del dev ens3 root
echo "set to normal"
break
fi
done
touch ended.info
On my client I generate a logfile which I plot with GNUPlot and I both calculate the avg uploadspeed on server and on the client. In this case 2740 kbit/s. Am I not using the tc tool correctly?
Image of my results generated with GNUPlot:
Upload speed
tc qdisc show dev ens3
gives me
qdisc pfifo_fast 0: root refcnt 2 bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1
which Im not able to delete with
sudo tc qdisc del dev ens3 root
Would be kind if someone could point me in the right direction and could explain why there is such a high upload rate, why there are frames coming through with a far higher throughput then shaped. Thank you.
Updated as adviced
Ok upload rate 80 kbit in tc gives me around 80*8 = 640 KiloBit/s. Still does not explain the fluctuation of the packet income