Total regression of model is greater than its component parts

Hello,

please i need help with this question.

I have model with 5 latent facotr (2 construct)

First construct is composed from 1 factor - 7 variable (TV)
Second construct is composed from 4 factor - 13 variable
(
PCH1- 4 variables
PCH2 - 3 variables
PCH3 - 3 variables
PCH4 - 3 variables)

If i create regression at SPSS Amos between factor
TV vs. PCH 1 - standardized regression weight = 0,64
TV vs. PCH 2 - standardized regression weight = 0,67
TV vs. PCH 3 - standardized regression weight = 0,39
TV vs. PCH 4 - standardized regression weight = 0,47

It's ok, but if i create model construct (Tv vs. PCH) where i use parcelling technigue and agregate variables from one factor to 1 variable.

Final regression is 0,802 (greater, than singles regression)

Is that possible?

Thanks for your time,
Z.