Predicting from averaged mixed effects models - getting huge CI using sjPlot

Question

I used MuMIn::model.avg to average several lme4::lmer models and I want to get and plot the predicted results from the averaged model. I tried using sjPlot::plot_model but I got unrealistic CI (ranging from about -50,000 to about +50,000 for a variable that has values ranging from 54-673).

I also tried to run similar models without the random factor, using simple linear models (lm) and got similar unrealistic CI.

Here is the code, I didn't add data because I don't think it is relevant but I can do it if required:

m1 <- lmer(Y ~ A*B + A*C + A*D + A*E + B*C + B*D + B*E + F + (1|RE), data = dt, REML = F)
dd <- dredge(m1, fixed = c("C","D","E","F"))
m1.avg <- model.avg(dd, fit = T)
plot_model(m1.avg, type="pred", terms = c("A","B")) #using different fixed effects still gives huge CI

# A,B - categorical variables with 2 and 3 levels, respectively
# C,D,E,F - continuous variables

I did find a way to solve this problem, but it makes calculation times much longer (and requires that I use scaled variables due to convergence issues which lmer does not have). Following this question I did everything using glmmTMB instead of lmer and then got the SE using predict and calculated the CI from it.