Instead you are using closing price, to calculate the EMA.
Use the EMA based on the HPF data stream.
This is only a recommendation.
Basically there are no big differences.
But since this was designed to determine long-term cycles,
EMA has a little too much noise.
Here you can see how to choose an alternative data source for EMA.
If 50 EMA have too much Lag for you use lower value,
for example 10 or even 5 period EMA.
The classic EMA with this period gives a lot of false signals,
EMA of HPF with period 5 almost never given false ones.