With transaction, it is a different case. If you were to compress it like that, the client would have to uncompress the transactions, every single time they validate it. Same thing with when someone mines a block. A side effect of this is the increase in the rate of transactions, which wouldn't be good, considering the tradeoff for it.
Compression will be very useful for some specific transactions, but for most it will not, so it is not worth it.
The transactions have to be decoded anyway, so pushing them through a bunzip2 should not be that dramatic.
In exchange, you could reduce fees to about a third and increase transactions per block 3 times (more or less, one would have to test with a significant number of transactions to get the average compression rate).
With fees going through the roof and tx/s being rather low, I don't see why increasing the tx rate would be bad.