CSV output uses previous point value when it should be a null

[ryantxu]

Bug report

X-Influxdb-Version →1.2.2

Steps to reproduce:

1. /query?q=DROP DATABASE csv_test
2. /query?q=CREATE DATABASE csv_test
3. /write?db=csv_test

example a=1,b=2 1465839830100400000 
example a=1 1465839830100500000 

Expected behavior:
I would expect the second point to have a null value for 'b' as it does in the json response:

"values":[["2016-06-13T17:43:50.1004Z",1,2],["2016-06-13T17:43:50.1005Z",1,null]]

Actual behavior:
JSON format gives the expected results, but CSV does not

/query?db=csv_test&q=SELECT * FROM example
Gives:

name,tags,time,a,b,c
example,,1465839830100400000,1,2
example,,1465839830100500000,1,2