the noise increases for half cells with one cycle formation in comparison with fresh half cells without any charge/discharge cycle. the connections are well.
Bubble (gas) evolution on the WE-surface may occur due to the fluctuations of dissolved (gas) concentration at the WE//electrolyte interface (bubble screening), shown, primarily, at low-Freq. in EIS.
Try to reduce, principally, the value of the high (cut-off) potential in your next (charge/discharge) cycle(s), and remeasure EIS (using the parametric value VDC=VOCP).
Remeasure EISb, after cycling; wait, for 1 day (of relaxation); remeasure EISc[1] of this, the same, cell. Check for, any possible, differences (between Z-spectra: EISb-EISc).
Also, what are the metal-substrates, e.g. the internal (metallic) current collectors in your button cells, used for your anodic (and cathodic) active materials ?
1. You must use the same, meaningful, (Z'-Z'') scales in your Nyquist plots (Capture44.JPG).
it might be, an additional issue for the noise shown in your system, the wrong implementation of Al as a metal-substrate for (all) your anode(s), either full, or half, Li-cells.
So, use Cu-foil (only, e.g. replace Al by Cu) as a metal-substrate for (all) your anode(s), both for:
'WE' (as anode-mat., here, in your case of half Li-cell), and