KNX Binding - Switch is working by OH but Status is not updating correctly

I installed Openhab 3 and activated the KNX binding.
I can also switch lights on and off via an item.
That means Openhab sends accordingly to the BUS.
What does not work, however, is if I turn the light on or off at the switch that Openhab receives the message.

For me this means that Openhab does not get the status of the Stauts group address.
I only see the information in the log file when I make changes directly from Openhab.

Does somebody has any idea?
If I trigger Telegrams by ETS, lamp also is working, but in the Logfile from OH I cannot see that something comes in.
The device is programmed as follows

UID: knx:device:CRU:25
label: KNX Device
thingTypeUID: knx:device
configuration:
pingInterval: 5
address: 0.0.0
readInterval: 5
fetch: false
bridgeUID: knx:ip:CRU
location: Keller
channels:
- id: sw_Arbeitszimmer
channelTypeUID: knx:switch
label: Deckenlicht Arbeitszimmer
description: null
configuration:
ga: 1/2/11+<1/2/12
UID: knx:ip:CRU
label: KNX/IP Gateway
thingTypeUID: knx:ip
configuration:
useNAT: false
readRetriesLimit: 3
ipAddress: 192.168.25.12
autoReconnectPeriod: 30
localIp: 192.168.25.14
type: TUNNEL
localSourceAddr: 0.0.0
readingPause: 50
portNumber: 3671

fixed.
After creating a new thing device everything worked

Please don’t use fake individual addresses (is address in the knx thing) but leave the parameter empty.
Please be aware that there is a fundamental difference between localSourceAddr (this is a fake individual address which must not be configured to any real device in ETS) and address (this is the real individual address of the knx device)
Please never use pingInterval that small, this is: test every 5 seconds, if the device is still alive. This will result in much traffic on knx bus.
Please never use readInterval > 0 if not absolutely needed. readInterval 5 is: send every 5 seconds a Read Request to any readable GA in this Thing (here: 1/2/12). This option is only for hardware which can’t send frequently or when the status changed. It’s very inefficient to pull data frequently.