rank sum test in R vs. MATLAB: exact p-value when ties are present

Question

Dear masters of R and MATLAB!

As documented, the rank sum test in R can not yield the exact p-value when the data has ties:

x=c(1,1,100,1000,10000);y=c(2,20,20,2000,20000);
result=wilcox.test(x,y,mu=0,paired=FALSE,
                   conf.level=0.95,exact=TRUE,alternative='greater')


Warning message:
In wilcox.test.default(x, y, mu = 0, paired = FALSE, conf.level = 0.95,  :
  cannot compute exact p-value with ties

Result

Wilcoxon rank-sum test with continuity correction

data:  x and y
W = 10, p-value = 0.7358
alternative hypothesis: true location shift is greater than 0

BUT MATLAB DOES IT:

x=[1,1,100,1000,10000]';y=[2,20,20,2000,20000]';
 [p,h,stats]=ranksum(x,y,'alpha',0.05,'method','exact','tail','right')

p = 0.714285714285714

I guess, MATLAB does an illegal thing (pMATLAB = 0.71, not 0.74), BUT i would need to force R to yield the exact p-value (pEXACT = pMATLAB = 0.71) for testing purposes.

Thank you :)