Why can't Holt Winters Seasonal recognize the seasonal component when forecasting a period

Question

I'm trying to predict a value based on historical data. The ts() function in R is loaded with 32 months. When I decompose this TimeSerie I will get the following result.

Printscreen Decompose TimeSeries

You can see a seasonal and trend.

When I run the R Script

install.packages(forecast)
install.packages(dplyr)
install.packages(TTR)
install.packages("fpp2")
library(forecast)
library(dplyr)
library(TTR)
library(ggplot2)
library(fpp2)
DummyDATA3$X<-c(58438489.43, 7760897.74, 6716940.83, 12419101.83,12061503.17,10489284.36,7543945.60,
                                 3933232.89, 9222001.26, 4587520.77, 5817801.17, 4832440.77, 27810414.64, 6597040.46,                                 
                                 5885012.59,7310623.51,4473150.98,5461631.24,6059878.47,4499952.91,4183396.09,4673311.51,
                                 4713036.50,2053582.50,4319487.49,1133142.80,1295063.09,929577.94,529379.77,323171.24)

DF<-DummyDATA$X
Gts<-ts(DF, start = 2016, frequency = 12)
decompose(Gts)
plot(decompose(Gts))
hw<-HoltWinters(Gts)
plot(hw)
forecast<-predict(hw,n.ahead = 24, prediction.interval = T, level = 0.95)
plot(hw,forecast)
forecast

this is the result:

Holt-Winters Filtering

Is there a way to get a better prediction with the seasonal component? e.g. I expected an increase at 2019.0.

I'm using the following data:

DummyData