understanding lambda overlay & autotune

[phantom_]

i have experimented with the official release 1.1.9.6 and autotune feature. please help me for clarifying some issues;

given an AFR reading at some RPM-MAP combination; lambda overlay function assigns some of this reading to one cell, and the remaining to the adjacent cell (trying to summarize what is documented in the help file). partitioning of this value depends upon max. interpolation value, and help file is clear for the interpolation strategy.

but why do we need interpolation for lambda overlay? tuner wants to see the fuel correction for every cell in the maps. isn't it that simple?

if the max. interpolation value is %100, i observe that the number of corrections increase in any datalog. does this mean do not interpolate around cells, just find the average AFR for every cell?

if no. of samples value increases, can we say that the confidence of the fuel corrections increases respectively?

autotune uses lambda overlay max interpolation value for the adjustment of fuel values. however; closest cell & precise cell strategy affects only one cell (as far as the help file says...), does this mean that there is no interpolationing for these strategies?

finally, in order to use autotune effectively; is it ok if the car is driven on a long, smooth road with no elevations, trying to generate data at different loads, different gears, varying speeds etc.? is this the appropriate method?

[Hondata]

Most of your questions you've answered yourself.

Without any interpolation at all, the fuel table turns jagged fairly quickly. That's why we don't normally use 'closest cell'. If you only want to see the fuel correction, then use the lambda overlay.

The best method is 'precise cell' on a load dyno. You can get reasonable results driving around, holding the rpm and varying the load. This is easier on the earlier vehicles which allow you to brake and throttle at the same time.