Same sample injections with huge difference in response, what could cause this?

Are the retention times ok?  As for the sampler, I would replace the seat with the needle.   Also, when was the last time the rotor seal or metering seal replaced?  

Could you show an example of a good vs bad run?  What model number is your auto sampler?

Yes the retention time is okay and the peaks and baseline look okay. The needle and needle seat were just changed on our last PM (February 14th). The autosampler model number is G7167B. I will have to check on the rotor seals or metering seal.

Our calibration comes in fine, the response is where it should be and we can get maybe one or two samples ran with not hiccup. Then we will get the fluctuation in the response. For our samples we test a sample, duplicate sample, spike sample and blank all with internal standard. The sample response would be ~850 where the duplicate sample response would be ~150 and the sample is prepared the exact same way.  

There may be two reason behind it 1) there may be carryover of sample from previous so set needle wash after every injection

2) check the analytical method , sometime the unvalidated method also shows this of problem

another suggestion : decrease the sample injection volume may be it can solve. 
it also happen once in my lab too and above method solve and onwards no problems seen yet