Upload & Sell: On
My speculation is that you metered on point, for each set of 3, as opposed to average and the point that you randomly selected resulted in large difference in exposure.
Even if you were on average, the light variation across the scene will cause difficulties because on a really wide angle your extreme left will be way differently lit than extreme right.
My strategy with ts17 is to meter with histogram in manual for middle normal exposure picture. Then leave the exposure constant across the 9 exposures. It works pretty well. But as I mentioned, using autometering (or resetting each set ) for each set has resulted in wide variations.