I am using R Studio. You did a very good job of aligning and normalizing the data, so it was easy to import the tab-delimited output files and do the R CCF stat function on the two sets.
plot(ccf(t_had[2],co2_ml[2],lag.max=360))
BTW, All the extra fine structure in the cross-correlation function is the matching of the somewhat periodic disturbances of dCO2 with the temperature fluctuations. That is a sign of further correlation. For example you would see that same structure if you did a cross-correlation of a sine curve with a cosine curve, albeit that would also show an obvious lag.
The overall envelope tells us about the impulse response, which is a disordered damped function.