Dear Stata Community,

First of all I wanted to say hi as I am new here and to express my gratitude to this forum, it has helped a lot.
At the moment I am writing my master thesis on buy and build acquisitions and I am having some trouble understanding how to use and interpret interacting variables.

To start of - I am using a panel dataset which is set per Strategy Year with independent var Return on Sales and independent variable distance (and controls but let's keep it simple). A strategy is composed of a platform company and the follow-on acquisitions done thereafter. On Rel Year 0 the platform company gets bought and once that is happened follow-on acquisitions might happen yes or no. If a follow-on acquistion happens the unconsolidated financials are summed and the distance between the companies is calculated (and if a second acq. is done it gets summed in the DistancePF). This looks as follows:
STRATEGY YEAR REL_YEAR sum_RoS_w POST sum_DISTANCEPF
1 2000 -1 .4 0 0
1 2001 0 .5 1 0
1 2002 1 .45 1 100
1 2003 2 .56 1 100
1 2004 3 .48 1 150
1 2005 4 .5 1 150
2 2003 -1 0.51 0 0
2 2004 0 0.52 1 30
2 2005 1 0.48 1 30
2 2006 2 0.47 1 30
2 2007 3 0.38 1 100
2 2008 4 0.52 1 100
etc.
Now I am staring with a simple regression between the depended and independent variables, but with an interaction term between Post and DistancePF. I am using reghdfe as I have to cluster my std err. on multiple clusters, but now I am in doubt if I have to use one # or a double ## and how to interpret the differences in results. My output looks like the below:
reghdfe sum_RoS_w Post Post#c.sum_DistancePFc, absorb(STRATEGY) vce(cluster STRATEGY NACE2#YEAR)
(dropped 74 singleton observations)
(MWFE estimator converged in 1 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gelbach & Miller applied.
note: 0b.Post#c.sum_DistancePFc omitted because of collinearity

HDFE Linear regression Number of obs = 2,372
Absorbing 1 HDFE group F( 2, 482) = 2.80
Statistics robust to heteroskedasticity Prob > F = 0.0620
R-squared = 0.5729
Adj R-squared = 0.4633
Number of clusters (STRATEGY) = 483 Within R-sq. = 0.0035
Number of clusters (NACE2#YEAR) = 744 Root MSE = 0.2418

(Std. Err. adjusted for 483 clusters in STRATEGY NACE2#YEAR)
------------------------------------------------------------------------------------------------------------------------
sum_RoS_w | Coef. Robust Std. Err. t P>|t| [95% Conf. Interval]
-----------------------+------------------------------------------------------------------------------------------------
Post | .0220763 .0135776 1.63 0.105 -.0046023 .048755
Post#c.sum_DistancePFc |
0 | 0 2.17e-16 0.00 1.000 -4.26e-16 4.26e-16
1 | .0000126 7.30e-06 1.72 0.085 -1.76e-06 .0000269
|
_cons | .0713037 .0113059 6.31 0.000 .0490888 .0935186
------------------------------------------------------------------------------------------------------------------------

Absorbed degrees of freedom:
-----------------------------------------------------+
Absorbed FE | Categories - Redundant = Num. Coefs |
-------------+---------------------------------------|
STRATEGY | 483 483 0 *|
-----------------------------------------------------+
* = FE nested within cluster; treated as redundant for DoF computation
the alternative with a double ## looks like this:

reghdfe sum_RoS_w Post##c.sum_DistancePFc, absorb(STRATEGY) vce(cluster STRATEGY NACE2#YEAR)
(dropped 74 singleton observations)
(MWFE estimator converged in 1 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gelbach & Miller applied.
note: 1.Post#c.sum_DistancePFc omitted because of collinearity

HDFE Linear regression Number of obs = 2,372
Absorbing 1 HDFE group F( 2, 482) = 2.80
Statistics robust to heteroskedasticity Prob > F = 0.0620
R-squared = 0.5729
Adj R-squared = 0.4633
Number of clusters (STRATEGY) = 483 Within R-sq. = 0.0035
Number of clusters (NACE2#YEAR) = 744 Root MSE = 0.2418

(Std. Err. adjusted for 483 clusters in STRATEGY NACE2#YEAR)
--------------------------------------------------------------------------------------------------------------------------------------------------------
sum_RoS_w | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-----------------------+--------------------------------------------------------------------------------------------------------------------------------
1.Post | .0220763 .0135776 1.63 0.105 -.0046023 .048755
sum_DistancePFc | .0000126 7.30e-06 1.72 0.085 -1.76e-06 .0000269
|
Post#c.sum_DistancePFc |
1 | 0 4.75e-21 0.00 1.000 -9.33e-21 9.33e-21
|
_cons | .0713037 .0113059 6.31 0.000 .0490888 .0935186
--------------------------------------------------------------------------------------------------------------------------------------------------------

Absorbed degrees of freedom:
-----------------------------------------------------+
Absorbed FE | Categories - Redundant = Num. Coefs |
-------------+---------------------------------------|
STRATEGY | 483 483 0 *|
-----------------------------------------------------+
* = FE nested within cluster; treated as redundant for DoF computation

My question now is which is better to use and how the output differs? If I understand it correctly the second one would be the correct way - but I am struggling to interpret the outcomes of this (specifically the interaction term).

Many thanks in advance to everybody reading this and taking the time to understand, help and answer me.

Kind regards,
George