pgmm and pvargmm producing different results in GMM estimations

Question

I'm trying to estimate a GMM model. I tried both the pgmm function from the plm package and the pvargmm function from the panelvar package. For some reason, the two packages do not generate the same results for the following regression model. I do not have the same problem when I do not introduce any non-linear terms in my specification (e.g., quits_rate.emp_rate.L). I'm quite confused about which package I should rely on and what's causing this difference? Any help would be much appreciated.

DC3 <- pgmm(investment_rate ~    
              lag(investment_rate) + 
              profit_rate.L + quits_rate.L
            + quits_rate.emp_rate.L | lag(investment_rate, 2:4)  + 
              lag(profit_rate.L, 1:4) + 
              lag(quits_rate.L, 1:4)+ 
              lag(quits_rate.emp_rate.L, 1:4)
            , effect = "twoways" , index=c("state", "year"), 
            data=pgmmm_df, collapse = T, robust=TRUE,
            transformation="d", model = "twosteps")

DC3.v2 <- pvargmm(
  dependent_vars = c("investment_rate"),
  lags = 1,
  predet_vars=c("profit_rate.L",
                "quits_rate.L",
                "quits_rate.emp_rate.L"
                ), 
  exog_vars=c("yr2003","yr2004","yr2005","yr2006"
              ,"yr2007", "yr2008", "yr2009","yr2010", "yr2011", "yr2012",
              "yr2013", "yr2014", "yr2015","yr2016", "yr2017", "yr2018", "yr2019"),
  transformation = "fd",
  data = pgmmm_df,
  panel_identifier = c("state", "year"),
  steps = c("twostep"),
  system_instruments = FALSE,
  max_instr_dependent_vars = 4,
  max_instr_predet_vars = 4,
  collapse = TRUE
)