Lambda inconsistencies

Hi! I have finished tuning the fuel map on my car, using a combination of steady state, ecumaster auto tune and log reviews with megalogviewerhd. All of this was done on the road. Checking cells in steady state, and using megalogviewerhd, shows everything to be correct, or at most off by 1%-2% on the log. I've got a couple things bothering me when watching the afr while driving though.

First is that it always starts out a bit rich when returning to idle. It will start at 0.85. then move up to .99 over the course of a few seconds. It is consistent and does this every time.

Second is that the lambda is not really following my target at all under normal driving conditions. I have attached a screenshot of normal driving. It has wild lean and rich spikes that I cant feel at all when I'm driving. It doesn't knock and I cant feel any misfires but these are still alarming to see. When doing wot acceleration tests it tracks the lambda target correctly. It also tracks correctly if I hold in the center of any cell and allow the lambda to stabilize. Closed loop control is off of course. The screenshot looks like I haven't tuned anything at all. I was expecting it to track the lambda target much more accurately.

What is going on here? Is this normal and not something to worry about? My o2 sensor is downstream from the head about 4 feet, if that matters. I have smoke tested the exhaust and intake and there are no leaks. I am not fully confident in the tunes accuracy because I haven't done this before and didn't use a dyno. My logs suggest I've done it right but if you think verifying the tune on a dyno would be a good use of time I can see if I can find one around me.

Thanks!

First, we don't worry too much about the lambda values in transient conditions as long as the engine is responsive. Slight rich on a throttle lift? Unless you have decel fuel cutoff (DCFO) active, this is pretty common as the manifold pressure goes lower, the fuel film on the port walls evaporates and is extra fueling you didn't request. ECU's with fuel film modeling can compensate for this, but it's not really a big deal (although it is wasting a bit of fuel if you are chasing fuel economy).

Realize that your Lambda readings are a bit delayed from the events you see, more at lower RPMs, and less at higher RPMs. So your lean spikes always follow throttle movements that have resulted in acceleration enrichment. Interestingly, your get a lean spike, then close to target / rich, followed by another lean spike. Changing the rate / decay values for that will probably affect the spikes, but if the engine responds well now, you can just ignore that.

In a log, I usually only focus on lambda values during steady-state or constant acceleration operation -- no throttle or map movement.