You can modify the original processing steps a little bit:

> First create a 1D file, AvsBcoding.1D, with 0's (at those TR's where
> neither condition A nor B occurred), 1's (at those TR's where
> condition A occurred), and -1's (at those TR's where condition B occurred).

Change the above step to: Create a file of 1's, 0's, and -1's at each tenth of a second.

Then use 1dUpsample to up-sample your original ROI time series, and then go ahead as usual (creating the interaction, run waver with specific duration, etc.). At the end down-sample the final regressors back to your original temporal resolution.

Gang