Use of emmean's option 'simple'

Hello.
I've found the package very useful but now I stumbled on something I consider a bug.

Say I have a model like this:
rt <- art(score ~ site*fsex + (1|patid), data = tofit)
were 'score' is a disease score, 'site' is a body region, 'fsex' is sex coded as factor and 'patid' is the patient ID.
I'm interested in the contrast of each site by sex. Say 'arm/male - arm/female'

If a use a linear model and the steps after
lm <- lmerTest::lmer(score ~ sitefsex + (1|patid), data = tofit)
LEMM <- emmeans(lm, ~ site:fsex)
pairs(LEMM, simple="fsex")
I get what I want. Whereas,
rt <- art(score ~ sitefsex + (1|patid), data = tofit)
EMM <- emmeans(rtlm, ~ sitefsex)
pairs(EMM, simple="fsex")
Will give me only NA's. I believe this is because LEMM has a 'site' column and an 'fsex' column, whereas in EMM these columns are combined into 'sitefsex'.
If there's a workaround for this I would be happy to know.

Reproduce the two emmeans and pairs calls from the issue, comparing the separate site and fsex columns with the combined sitefsex result. Trace how simple="fsex" is interpreted for the ART model and confirm the expected contrast output; done means the behavior is corrected or a verified workaround is documented.