Well, it sounds like you have proceeded carefully. Since the SVI fit is certainly not guaranteed to lie between the market bid-ask, it seems like you were right in your observation that it's just poor performance. The Fengler method fit is likely much better, and fomisha just posted another one to check out.1. These are european optionsSome possible issues beyond simply out-of-sync or bad market data:
1. Since this is a single-name stock, what is the exercise style and how are you converting to Euro-style if it is American-style?
2. Are you using exclusively out-of-the-money options to calculate IV's? (I would, so puts for K < S0, calls for K > S0).
3. Assuming Euro-style or a reasonable conversion to that exercise style, are you then enforcing put-call parity and, if so, how? The CBOE has a method in the "VIX white paper", which will yield an option implied forward price. I would calculate that and compare to the futures price. (BTW, from where are you getting that futures price?) If they are different, try re-running the whole analysis with that option-implied forward price (new IV's, new SVI run, etc).
2. Yes, I'm using OTMs exclusively
3. Although I havent enforced put call parity conditions, I checked and PCP holds for the prices above if i include transaction costs and spreads
As far as the data is concerned, these data points are in sync and have been taken from a TBT feed