Other than annealing temperature (too high, in my opinion), there is nothing obviously wrong with your parameters. You have to realize that there is no set of parameters that works for all protein alignments. For example, here is a set of parameters that often works form me, but sometimes it does not.
lset rates=invgamma nst=6 ngammacat=4; prset statefreqpr=fixed(empirical); mcmcp nruns=2 ngen=1000000 printfreq=1000 samplefreq=100 starttree=random; mcmcp relburnin=yes burninfrac=0.25 mcmcdiagn=yes diagnfreq=1000 stoprule=yes stopval=0.0005; mcmcp nchains=4 nswaps=2 allchains=yes checkfreq=1000 savebrlens=yes temp=0.1;
You need to find best parameters for your alignment. There is a program for that - see here or ModelFinder in IQ-TREE. You can also try doing it by doing the permutations manually. For example, you do the run with options listed above. Next time change
rates=gamma and so forth. Beware that it typically takes much longer to do it manually as opposed to automatic selection.
By the way, I would try
temp from one of the following values
[0.05, 0.075, 0.1, 0.125] as
0.25 is probably too much heating.