URGENT: non-consecutive lags (excluded lags) in VAR/ VEC

[musikmachtgluecklich]

Dear experts,

is there any problem in skipping lags in a VAR/ VEC?
I estimated models like a VAR with lags at order 1, 5, 6, 23, 36, but not the ones inbetween, and similar models, encouraged by the fact that the eviews Help explicitely instructs you how to do that.
My only concern is that i didn't find a single paper where economists would use non-consecutive lags for VEC or VAR estimation (like me).

How I got there:
I came up with those lags since I'm working with daily data, the samle size is 1000 data points, and when it comes to lag lenght crieria, the LR statistic chooses huge lag orders (e.g. 47 out of a specified max. of 50, 75 out of max 100) while SIC and HQ choose 1 lag only.
So I estimated a VAR(1), had a look at the correlograms and autocorrelation LM tests, included all the lags that seemed significant. Finally, i performed lag exclusion tests on this huge model with all the included lagged terms and according to the lag exclusion test results, started eliminating lags one after the other, beginning with those with the highest p- values in the lag exclusion test, until all remaining lags were significant at the 5% level.

is that a good way (or at least an acceptable way) to proceed?
does anybody know of any serious studies were researchers proceeded likewise?

Thank you so much!
Kerstin