This site never mentions picking specific "fruits" to save seeds from, only "plants".
http://www.kdcomm.net/~tomato/...
http://kdcomm.net/~tomato/gene...
It does go into why it takes so many generations to "fix" one or more traits.
And why it is so much faster to select for recessive traits than dominant traits.
My guess is that if you don't see definite improvement after 2-3 years, fruit shape is either controlled by many genes with partial dominance, or environmental effects.
It also gives good advice about how to cross two plants despite tomato's tendency to self-pollinate. (Remove the anther cone from the female parent bloom
before the flower opens, aiming for 24-36 hours before the pollen would have dropped.)
It sounds to me as if the practices that probably pre-date Mendel are still just about the best we can do without gene guns and modern lab techniques like Agrobacterium plasmids or CRISPR.
Remove the poorer plants.
Protect your seed plants from undesirable cross-pollination.
Collect seeds from the very best plants.
Repeat for at least 8 generations.
After that, still pay attention and keep "roguing out" throwbacks and undesirable plants.
P.S. While you're playing the numbers game with 64, 128 or 256 plants, be sure to taste each plant! Wouldn't it be a shame if you bred some wonderful-tasting tomato, or one that cured baldness and depression, but never realized the fact and composted it! One like that would be worth taking cuttings and offering to serious research labs for DNA sequencing.