The growth rate of Rwandan manufacturing companies - a puzzle

This post was going to be about balance in the industrial structure of the Rwandan economy, but my results have produced something that I can't explain, probably partially due to my error.  But I think there might be something going else on too, beyond any of my mistakes.

Data: http://www.enterprisesurveys.org/

The graph shows the growth rates of Rwandan manufacturing companies over the period 2009-2011, when the companies were operating at the start and end dates and didn't grow faster than 100% per year.  The average rate of annual growth is -11.6 percent, which I am finding difficult to believe.  I have subtracted the average inflation rate (11.4%) from the data, but perhaps the data is already adjusted for inflation and it should be added back.  But the average Rwandan manufacturer that survived the period would then still have had a negative growth rate.  The Rwandan economy was growing rapidly during the period, hence my continued confusion.

The results could be explained by very rapid growth in very few companies, or by new companies creating a lot of value.  These explanations may work, but still...

Here's the Stata code:
use "C:\Rwanda-2011-full data-999.dta", clear  //From http://www.enterprisesurveys.org/
gen annual_sales_perc_growth=100*((d2/n3)^(1/3)-1) if d2>=0&n3>=0
replace annual_sales_perc_growth=annual_sales_perc_growth-11.4
gen manuf_sector=(a4a==1) if annual_sales_perc_growth<=100
hist annual_sales_perc_growth if manuf_sector==1, percent xtitle("Annual percentage growth rate, 2009-2011") ytitle("Percentage of companies")
mean annual_sales_perc_growth if manuf_sector==1

//More carefully, using the survey structure
svyset idstd [pweight=median_weights], strata(strata) singleunit(certainty)
replace manuf_sector=(a4a==1&annual_sales_perc_growth&<=100&d2>=0&n3>=0)
svy, subpop(manuf_sector): mean annual_sales_perc_growth