diagnostics question - good Gelman, bad Geweke

[amulcan]

I have 3 tracers, 4 sources, about 100 samples/mixtures, and no discrimination factors. I have run multiple iterations of the same/similar models, most of which have converged as indicated by good Gelman and Geweke diagnostics. Based on these I have a pretty good idea of which sources are influencing the mixtures in my study area. I run all models using the "extreme" chain length option. Despite the success of similar models, one of my models in particular yields good Gelman, but bad Geweke diagnostics, which are:

Out of 489 variables Gelman - should be < 1.05 65 > 1.01 2 > 1.05 0 > 1.1

Geweke - we expect 5% of variables to be outside +/- 1.96 Chain 1 50 (10.2% of variables outside +/- 1.96) Chain 2 86 (17.6% of variables outside +/- 1.96) Chain 3 73 (14.9% of variables outside +/- 1.96)

Is this model converged? In other posts I have seen it suggested that the Gelman diagnostic is more important for determining model convergence than the Geweke, but I'm not sure of this.

Any help is appreciated, thanks in advance!