Okay two things...
1) Your Quota is set to <Distributed>false</Distributed>.
By default your Apigee instance runs on two separate Message Processors (the servers that do the heavy lifting). This means that each MP will count and with a Quota of 6 you effectively have 6 * 2 servers = 12.
2) Your Quota is Distributed but Asynchronous in the second example.
If you don't set <Distributed> and <Synchronous> to false, Apigee will share Quota counts by checking in with the central data server. There will always be some lag with this, but you have set your AsynchronousConfiguration to check in with the central server every 20 seconds or every 5 messages, so you could, count up to 5 on each MP processor before checking in with the other servers.
Keep in mind that in a distributed processing model like Apigee you will never get an absolutely precise number because even with Distributed set to true and Asynchronous set to false there will always be some lag with the servers talking to each other.
Also, you might want to strip out the request.header.quota_count and other request.header variables -- if I passed a number (say 100000) as a header like
quota_count: 100000
Apigee will use the 100000 rather than your value of 1 (it uses the referenced variables and rolls back to the default value only if the reference is NULL).
And... you probably want to add an <Idnetifier ref="client_ip"> or something otherwise the quota is global to all users. See the Apigee variables reference at http://apigee.com/docs/api-services/api/variables-reference for the variables that are available in every flow.
