Preventing ggplot2 from printing a vector in legend

Question

I've used ggbiplot to produce a biplot. I want points of different colours for different factor levels and the size of points to represent sample size. The problem that's driving me nuts is that part of the size vector c(1,2,2,1,3,2,2,1,3,3,2,1,1,1,2,3,2,2,3,1 is printed in the legend:

I used:

library(ggbiplot)
pplot <- function(pcaMatrix, cultures, sample.size = NA, FIRST=1, SECOND=2, ELLIPSE.PROB=0.68, SCALE = FALSE, LABELS=NULL){
  PCA <- prcomp(na.omit(pcaMatrix), scale. = SCALE)
  plot <- ggbiplot(PCA, obs.scale = 1, var.scale = 1, labels=LABELS, ellipse = {Ellipse.prob>0}, groups=cultures, circle = FALSE, choices=c(FIRST,SECOND), ellipse.prob=ELLIPSE.PROB)
  plot <- plot + geom_point(aes_string(colour=cultures, size = sample.size)) #+ scale_size_identity() 
  plot <- plot + scale_color_discrete(name = '')
  plot <- plot + scale_size_continuous(labels = c("200+", "500+", "1000+", "2000+", "4000+"))
  plot <- plot + theme(legend.direction = 'horizontal', legend.position = 'top')
  plot <- plot + theme(aspect.ratio=1)
  print(plot)
}

pplot(grav, grav.cultures, grav.sample, 1,2, ELLIPSE.PROB=0)

where

> head(grav)
      Microlithisme Substrat     BpBG Crans  Gm   PD   F
    1          25.0     29.9 61.94226   0.5 0.3 27.0 0.3
    2          44.8     29.3 20.93023   2.1 0.0 47.9 0.0
    3           7.0     66.2 57.58929   0.5 0.2  5.8 0.2
    4          13.2     12.9 55.41126   6.6 0.0 40.0 0.2
    5           2.0      9.0 75.03045   0.1 0.0  6.9 0.0
    6          16.4     17.3 51.36519   0.9 0.2 20.9 0.1

> head(grav.cultures)
[1] GR GR GR GR GR GR
Levels:  AUR AZI EG EGL GR GR.Italy MAG SOL

The actual sample size is modified to control manually the size of points:

> head(grav.sample)
[1]  360  986  599  457 1222  899

grav.sample[grav.sample>=200 & grav.sample < 500] <- 1
grav.sample[grav.sample>=500 & grav.sample<1000] <- 2
grav.sample[grav.sample>=1000 & grav.sample<2000] <- 3
grav.sample[grav.sample>=2000 & grav.sample<4000] <- 4
grav.sample[grav.sample>=4000 & grav.sample<8000] <- 5

> head(grav.sample)
[1] 1 2 2 1 3 2

I really just can't figure this one out. Why is part of the size vector printing in my legend?