在data.table的“j”参数中没有得到相同的结果

我从data.table的“j”参数中调用一个函数,如果我直接调用它,得到的结果将不同。这似乎与 my function called in data.table j not returning expected results

get.lower.bound <- function (x) {
  rex <-
    regexpr (pattern = "((?<lower>[\\-+\\d*\\.,]*)%\\s*<\\s*)?X(\\s<\\s(?<uppper>[\\-+\\d\\.,]*)%)?",
             text = x,
             perl = TRUE)
  lower_bound <-
    substring(
      text = x,
      first = attr(rex, "capture.start")[2],
      last = attr(rex, "capture.start")[2] + attr(rex, "capture.length")[2] -1
    )
  lower_bound
}

dat <- data.table(
  A = c('1% < X < 2%', '4% < X',      'X < 8%' ),
  B = c('2% < X < 3%', '5% < X < 6%', '8% < X < 9%' ),
  C = c('3% < X < 4%', '6% < X < 7%', 'X < 10%' )
)

get.lower.bound(dat[1,'A'])                   # this returns 1; as I expect
get.lower.bound(dat[2,'A'])                   # this returns 4; as I expect
get.lower.bound(dat[3,'A'])                   # this returns a blank string; as I expect
dat[i = 1, j = .(lb1 = get.lower.bound(A))]   # this returns a data table of just one cell with 1 in it; as I expect
dat[i = 1:3, j = A]                           # this returns a character vector with the original strings in it; as I expect
dat[i = 1:3, j = .(lb1 = c(A))]               # this returns a data table with the original strings in it; as I expect
dat[i = 1:3, j = .(lb1 = get.lower.bound(A))] # this returns "1% <", "4% <", "X < 8"; but I expect a "1", "4" & ""

为什么最后一句话没有给我我想要的?我需要做什么才能得到我想要的?